Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
UNIVERSITÉ DE MONTRÉAL
RÉACTIONS HOMOGÈNES EN PHASE GAZEUSE
DANS LES LITS FLUIDISÉS
JEAN-PHILIPPE LAVIOLETTE
DÉPARTEMENT DE GÉNIE CHIMIQUE
ÉCOLE POLYTECHNIQUE DE MONTRÉAL
THÈSE PRÉSENTÉ EN VUE DE L’OBTENTION
DU DIPLÔME DE PHILOSOPHIAE DOCTOR (Ph.D.)
(GÉNIE CHIMIQUE)
JUIN 2010
© Jean-Philippe Laviolette, 2010.
UNIVERSITÉ DE MONTRÉAL
ÉCOLE POLYTECHNIQUE DE MONTRÉAL
Cette thèse intitulée:
RÉACTIONS HOMOGÈNES EN PHASE GAZEUSE DANS LES LITS FLUIDISÉS
présentée par : LAVIOLETTE, Jean-Philippe
en vue de l’obtention du diplôme de : Philosophiae Doctor
a été dûment acceptée par le jury d’examen constitué de :
M. LEGROS Robert, Ph.D., président
M. PATIENCE Gregory S, Ph.D., membre et directeur de recherche
M. CHAOUKI Jamal, Ph.D., membre et codirecteur de recherche
Mme LA MARCA Concetta, Ph.D., membre et codirecteur de recherche
M. FRADETTE Louis, Ph.D., membre
M. HEMATI Mehrdji, Doct.Ing., membre
iii
REMERCIEMENTS
Je remercie mes directeurs de thèse, Jamal Chaouki et Gregory Patience, pour m’avoir accueilli
dans leur équipe de recherche, pour l’aide qu’ils m’ont apportée, pour leur patience et pour leur
encouragement. Leurs nombreuses critiques m’ont été très précieuses dans mes travaux.
Je remercie également ma co-directrice de thèse, Concetta La Marca, qui m’a accueilli chez
E.I. du Pont de Nemours afin que je puisse mettre en pratique mes connaissances dans un
contexte industriel. Nos discussions sur la modélisation des réactions m’ont été très utiles.
Je tiens à remercier Jean Huard, Daniel Dumas, Carol Painchaud, Robert Delisle, Gino Robin et
Martine Lamarche sans qui, je n’aurais pas pu réaliser ce travail.
Le déroulement de mon doctoratn’aurait pas été aussi agréable sans avoir rencontré Farhad Azar,
Ali Shekari, Babak Esmaeili, Rouzbeh Jafari, Mania, Jocelyn Doucet, Jonathan Bouffard et
Olivier Dubé.
Je remercie ma famille qui m’a soutenu pendant ce travail et plus spécifiquement mes parents qui
ont toujours eu mon bien-être à cœur. C’est en grande partie grâce à eux si je me trouve là où je
suis présentement.
Finalement, je remercie ma conjointe, Florence, qui a quitté Marseille pour m’accompagner dans
cette aventure qui devait à l’origine durer… 3 ans.
iv
RÉSUMÉ
Cette thèse présente une étude sur les réactions homogènes en phase gazeuse dans les lits
fluidisés. L’objectif principal est de développer des outils de modélisation et de caractérisation
des réactions homogènes en phases gazeuses dans les réacteurs à lit fluidisé.
Minimiser l’occurrence de réactions homogènes dans les lits fluidisés est un enjeu majeur
puisqu’elles peuvent causer une baisse de production suite à la dégradation des réactifs et des
produits désirés. Par ailleurs, ces réactions peuvent représenter un risque pour la sécurité
lorsqu’elles sont rapides et exothermiques comme c’est le cas pour les réactions d’oxydations.
Dans la région du lit fluidisé, la fraction de solides augmente le taux de recombinaison des
radicaux libre et évite l’initiation de réactions homogènes ou réduit leur vitesse de réaction. Il
existe toutefois des régions de haute porosité où des réactions homogènes peuvent être produites
par de fortes concentrations de réactifs combinés à des températures et pressions élevées
(caractéristiques de nombreux procédés incluant les procédés d’oxydation). Dans ces régions, le
taux de recombinaison des radicaux libres peut être insuffisant pour empêcher l’initiation de
réactions en phases gazeuses. Dans un réacteur à lit fluidisé, les régions de faible fraction de
solides incluent principalement:
• La zone du distributeur et des injecteurs dans le lit fluidisé
• La phase bulle dans le lit fluidisé
• Les régions en aval du lit fluidisé (zone de désengagement, cyclone, etc.)
Dans la première partie de ce travail, la combustion en lit fluidisé du méthane, éthane, propane et
n-butane avec de l’air a été étudiée en mode d’injection séparée. Un lit fluidisé de particules
inertes de sables a été opéré dans le régime à bulle et à des températures intermédiaires :
923 K ≤ TB ≤ 1123 K. Pour l’éthane, le propane et le n-butane, la combustion avait lieu
principalement dans la région de désengagement lorsque la température du lit fluidisé était
inférieure à T1 = 923 K. D’un autre côté, la combustion se produisant entièrement dans une
distance de 0.2 m de l’injecteur lorsque la température du lit était supérieure à T2 = 1073 K. Pour
v
le méthane, les valeurs de T1 et T2 étaient significativement plus élevées avec des mesures
respectives de 1023 K et 1123 K. La combustion dans le lit fluidisé a été modélisée avec
précision à l’aide d’une cinétique globale de première ordre et une modèle hydrodynamique
composé de deux réacteurs pistons en séries : un qui caractérisait la région près de l’injecteur et
un autre pour la région de lit fluidisé en aval. La cinétique globale de réaction a été mesurée pour
chacun des hydrocarbures. Pour les n-alcanes C2 à C4, la cinétique globale était caractérisé par
une seule expression de type Arrhenius, alors que la cinétique du méthane était significativement
plus lente. Les propriétés de la région de l’injecteur causaient soit une conversion importante
dans cette région, soit un délai d’auto-allumage dans le lit fluidisé. La conversion des réactifs
près de l’injecteur augmentait avec une hausse de la température du lit fluidisé et diminuait avec
une hausse de la vitesse d’injection du gaz à l’extrémité de l’injecteur. Afin de prendre ce
phénomène en compte dans le modèle de réaction, une analogie a été effectuée avec le concept de
temps d’induction. Ainsi, la longueur du réacteur piston caractérisant la région de l’injection a été
corrélée à la température du lit fluidisé et la vitesse d’injection avec une expression de type
Arrhenius.
Dans la seconde partie de cette thèse, la combustion du propane dans la zone de désengagement
d’un lit fluidisé de sable a été étudiée. La température du lit fluidisé a été maintenue faible entre
818 K et 923 K afin que la combustion se produise essentiellement en aval du lit fluidisé. De
plus, une vitesse superficielle de gaz deux fois supérieure à la vitesse minimale de fluidisation a
été utilisée pour minimiser l’emportement de solides. Durant les expériences, la zone de
désengagement a été caractérisée par des mesures simultanées de flux de solides, de composition
chimique de la phase gazeuse, de température et de pression. L’auto-allumage de la mixture
propane/air a été observé à seulement 0.06 m de la surface du lit pour des températures de lit
fluidisé supérieures à 833 K. Six mécanismes de microcinétique ont été utilisés afin de reproduire
le profil axial de conversion du propane: tous les modèles ont sous-estimé le taux de réaction
dans la zone de désengagement. Or, lorsque le modèle considérait le H2O2 produit lors de la
combustion dans le lit fluidisé, la vitesse de réaction calculée dans la zone de désengagement
étaient significativement plus élevée et causait une meilleure concordance entre le modèle et les
résultats expérimentaux.
vi
La troisième partie de ce travail relate le développement d’une nouvelle méthode spectroscopique
pour la mesure quantitative et simultané de la fraction de solides (1-ε) et de la composition
chimique dans la phase gazeuse (Yi) dans un système gaz/solide. La méthode a été utilisée avec
une FT-IR connecté à une sonde à fibres-optiques à laquelle des mesures en temps réel et in situ
pouvaient être effectuées. L’effet de (1-ε) et Yi sur le spectre d’absorbance mesuré étaient additifs
et pouvaient être calibrés de manière indépendante. Des mesures ont été effectuées avec des
mixtures alkane/azote et deux types de particules: du sable et du catalyseur FCC. La fraction
volumétrique de l’hydrocarbure et (1-ε) ont été variées entre des valeurs respectives de
1.8 - 10.1 mol% et 0 - 0.45. Les erreurs relatives sur la mesure de Yi étaient toujours inférieures à
6% and l’erreur augmentait avec une baisse de l’intensité du rayon infrarouge. Une preuve de
concept a été effectuée dans un lit fluidisé: la sonde à fibres-optiques a été utilisée pour mesurer
la fraction volumétrique d’un traceur gazeux dans la phase bulle et émulsion.
Les résultats ont également démontré que la méthode est limitée par deux paramètres:
• L’intensité du rayon infrarouge transmis dans l’échantillon et récolté par le détecteur du
spectromètre.
• La vitesse de mouvement des solides et la fréquence de modulation du spectromètre.
Un signal infrarouge faible augmente l’erreur sur les mesures instantanées. L’erreur relative sur
la mesure peut toutefois être significativement diminuée en-dessous de 5% si une moyenne
temporelle est effectuée. L’intensité du rayon infrarouge pourrait être accrue en:
• Augmentant l’intensité de la source IR
• Optimisant le transfert du rayon IR à l’interface spectromètre/sonde à fibres-optiques
• Optimisant le transfert du rayon IR l’extrémité de la sonde
• Augmentant la sensibilité du détecteur
vii
Le mouvement des particules solides produit du bruit dans le spectre d’absorbance mesuré de
sorte que la fréquence de modulation doit être suffisamment élevée afin de déplacer le bruit vers
les basses fréquences à l’extérieur des longueurs d’ondes d’intérêts.
viii
ABSTRACT
This thesis presents a study on homogeneous gas-phase reactions in fluidized beds. The main
objective is to develop new tools to model and characterize homogeneous gas-phase reactions in
this type of reactor.
Minimizing the risk of gas-phase reactions in fluidized beds is critical since they may cause a
decrease in production by degrading reactants and desired products. Furthermore, they may
constitute a safety concern in the case of fast and exothermic reactions such as oxidation
reactions. In the fluidized bed region, the high solids volume fraction quenches free radicals,
which minimizes gas-phase reactions. However, fluidized bed reactors are characterized by
several regions of low solids volume fractions where gas-phase reactions may occur due to high
temperatures, pressures and reactants concentrations. In these regions, the rate at which free
radicals are quenched may be insufficient. In a fluidized bed reactor, the regions of low solids
volume fraction mainly include:
• The regions close to the distributor and the injectors
• The bubble phase
• The regions downstream of the fluidized bed surface (freeboard, cyclones, etc. )
Chemical processes in fluidized beds should be designed in order to prevent the initiation of
homogeneous gas-phase reactions in these regions.
In the first part of this work, the non-premixed combustion of C1 to C4 n-alkanes with air was
investigated inside a bubbling fluidized bed of inert sand particles at intermediate temperatures:
923 K ≤ TB ≤ 1123 K. For ethane, propane and n-butane, combustion occurred mainly in the
freeboard region at bed temperatures below T1 = 923 K. On the other hand, complete conversion
occurred within 0.2 m of the injector at: T2 = 1073 K. For methane, the measured values of T1
and T2 were significantly higher at 1023 K and above 1123 K, respectively. The fluidized bed
combustion was accurately modeled with first-order global kinetics and two one-phase PFR
ix
models in series: one PFR to model the region close to the injector and another to represent the
main fluidized bed body. The measured global reaction rates for C2 to C4 n-alkanes were
characterized by a uniform Arrhenius expression, while the global reaction rate for methane was
significantly slower. Reactions in the injector region either led to significant conversion in that
zone or an autoignition delay inside the main fluidized bed body. The conversion in the injector
region increased with rising fluidized bed temperature and decreased with increasing jet velocity.
To account for the promoting and inhibiting effects, an analogy was made with the concept of
induction time: the PFR length (bi) of the injector region was correlated to the fluidized bed
temperature and jet velocity using an Arrhenius expression.
In the second part of this work, propane combustion experiments were conducted in the freeboard
of a fluidized bed of sand particles at temperatures between 818 K and 923 K and at superficial
gas velocity twice the minimum fluidization velocity. The freeboard region was characterized by
simultaneous measurements of solids flux, chemical composition, temperature and pressure.
Autoignition was only recorded within 0.06 m of the bed surface at temperatures greater
than 833 K. Propane conversion predicted by six different microkinetic mechanistic models were
compared to the experimental measurements: all six models underestimated the reaction rate
above the bed surface. However, accounting for the production of H2O2 during in-bed
combustion significantly increased the calculated reaction rates and resulted in a better agreement
between predicted and measured propane conversion.
In the third part of this work, a novel spectroscopic method was developed to measure
quantitatively and simultaneously solids volume fraction (1-ε) and gaseous species composition
(Yi) in a gas/solid system. The method was comprised of an FT-IR coupled to a fibre-optic probe
that could perform real-time and in-situ measurements of absorbance. The effect of (1-ε) and Yi
on the absorbance spectra were additive and could be independently calibrated. Experiments were
conducted with alkane/nitrogen mixtures and two types of particles: sand and FCC. Fuel mole
fractions and (1-ε) were varied between 1.8 - 10.1 mol% and 0 - 0.45, respectively. The relative
errors for Yi time-averaged measurements were below 6% and the error increased significantly
with decreasing beam intensity. A proof of concept for a novel application in fluidized beds was
x
also completed: the fibre-optic probe was used to measure the molar fraction of a tracer gas
inside the emulsion and bubble phases during gas tracer experiments.
xi
TABLE DES MATIÈRES
REMERCIEMENTS ..................................................................................................................... III
RÉSUMÉ ........................................................................................................................... IV
ABSTRACT ........................................................................................................................ VIII
TABLE DES MATIÈRES ............................................................................................................ XI
LISTE DES TABLEAUX .......................................................................................................... XIV
LISTE DES ANNEXES .............................................................................................................. XV
LISTE DES FIGURES ............................................................................................................... XVI
LISTE DES SIGLES ET ABRÉVIATIONS .......................................................................... XVIII
INTRODUCTION ........................................................................................................................... 1
CHAPITRE 1 REVUE DE LITTÉRATURE ............................................................................. 4
1.1 COMBUSTION EN LIT FLUIDISÉ .............................................................................. 4
1.2 COMBUSTION DANS LA ZONE DE DÉSENGAGEMENT ...................................... 7
1.3 CARACTÉRISATION DES SYSTÈMES MULTIPHASIQUES ................................ 10
CHAPITRE 2 PRÉSENTATION DES ÉTAPES DE TRAVAIL ............................................ 14
CHAPITRE 3 MÉTHODOLOGIE ........................................................................................... 16
3.1 EXPÉRIENCES DE COMBUSTION EN LIT FLUIDISÉ .......................................... 16
3.2 EXPÉRIENCES DE COMBUSTION DANS LA ZONE DE
DÉSENGAGEMENT ................................................................................................................ 17
3.3 MODÉLISATION DE LA COMBUSTION EN LIT FLUIDISÉ ................................ 19
3.4 MODÉLISATION DE LA COMBUTION DANS LA ZONE DE
DÉSENGAGEMENT ................................................................................................................ 20
3.5 DÉVELOPMENT D’UNE MÉTHODE POUR LA MESURE SIMULTANÉE
DE LA FRACTION DE SOLIDES ET DE LA COMPOSITION CHIMIQUE DE LA
PHASE GAZEUSE ................................................................................................................... 21
xii
CHAPITRE 4 COMBUSTION EN LIT FLUIDISÉ DES N-ALKANES C1-C4 .................... 25
4.1 PRÉSENTATION DE L’ARTICLE ............................................................................. 25
4.2 FLUIDIZED BED COMBUSTION OF C1-C4 N-ALKANES .................................... 26
4.2.1 Abstract ..................................................................................................................... 26
4.2.2 Introduction ............................................................................................................... 27
4.2.3 Experimental ............................................................................................................. 30
4.2.4 Results and discussion ............................................................................................... 31
4.2.5 Conclusions ............................................................................................................... 42
4.2.6 Nomenclature ............................................................................................................ 43
4.2.7 References ................................................................................................................. 44
CHAPITRE 5 COMBUSTION DU PROPANE DANS LA ZONE DE
DÉSENGAGEMENT D’UN LIT FLUIDISÉ ............................................................................... 48
5.1 PRÉSENTATION DE L’ARTICLE ............................................................................. 48
5.2 GAS-PHASE COMBUSTION IN THE FREEBOARD OF A FLUIDIZED BED ...... 49
5.2.1 Introduction ............................................................................................................... 49
5.2.2 Experimental ............................................................................................................. 52
5.2.3 Kinetic modeling in the freeboard ............................................................................. 52
5.2.4 Results and discussion ............................................................................................... 54
5.2.5 Conclusions ............................................................................................................... 67
5.2.6 Nomenclature ............................................................................................................ 67
5.2.7 References ................................................................................................................. 68
CHAPITRE 6 NOUVELLE MÉTHODE DE SPECTROSCOPIE POUR LA MESURE
SIMULTANÉE ET QUANTITATIVE DE LA COMPOSITION CHIMIQUE DE LA
PHASE GAZEUSE ET DE LA FRACTION DE SOLIDES ........................................................ 72
6.1 PRÉSENTATION DE L’ARTICLE ............................................................................. 72
xiii
6.2 SIMULTANEOUS QUANTITATIVE MEASUREMENT OF GASEOUS
SPECIES COMPOSITION AND SOLIDS VOLUME FRACTION IN A GAS/SOLID
FLOW ....................................................................................................................................... 73
6.2.1 Introduction ............................................................................................................... 73
6.2.2 Experimental ............................................................................................................. 77
6.2.3 Results and discussion ............................................................................................... 79
6.2.4 Application in fluidized beds - chemical composition measurement in the
bubble phase .......................................................................................................................... 91
6.2.5 Limitations of the technique and future work ........................................................... 94
6.2.6 Conclusions ............................................................................................................... 95
6.2.7 Acknowledgment ...................................................................................................... 96
6.2.8 Nomenclature ............................................................................................................ 96
6.2.9 References ................................................................................................................. 97
CHAPITRE 7 DISCUSSION GÉNÉRALE ........................................................................... 101
CONCLUSION ......................................................................................................................... 106
BIBLIOGRAPHIE ...................................................................................................................... 109
ANNEXES ............................................................................................................................... 119
xiv
LISTE DES TABLEAUX
Table 4.1: Measured T1 and T2 for C1 to C4 n-alkanes .................................................................. 36
Table 4.2: Jet penetration depth correlations ................................................................................. 38
Table 4.3: Parameter bi at the different bed temperatures (TB) ...................................................... 39
Table 5.1: Average gas residence time in fluidized bed ................................................................ 65
Table 6.1: Simultaneous measurement of YCH4 and (1-ε) .............................................................. 89
xv
LISTE DES ANNEXES
ANNEXE 1 – ARTICLE PRESENTÉE À LA CONFÉRENCE CHEMCON 2006 (INDE) .... 119
ANNEXE 2 – ARTICLE PRESENTÉE À LA CONFÉRENCE IFSA 2008 ............................. 129
ANNEXE 3 – ARTICLE PRÉSENTÉ À LA CONFÉRENCE FBC 2009 (CHINE) ................. 146
ANNEXE 4 – ARTICLE PRESENTÉ À LA CONFÉRENCE WCCE8 2009
(MONTRÉAL, CANADA) ......................................................................................................... 157
ANNEXE 5 – ARTICLE PRESENTÉ À LA CONFÉRENCE FLUIDIZATION XIII
(CORÉE DU SUD, 2010) ........................................................................................................... 167
ANNEXE 6 – ARTICLE PRÉSENTÉ À LA CONFÉRENCE FLUIDIZATION XIII
CONFERENCE (CORÉE DU SUD, 2010) ................................................................................ 179
ANNEXE 7 – CHAPITRE 9 DU « WILEY HANDBOOK OF COMBUSTION »
(VOLUME 5) (2010) .................................................................................................................. 191
ANNEXE 8 – ARTICLE PUBLIÉ DANS LE JOURNAL « APPLIED CATALYSIS A:
GENERAL » (2010) .................................................................................................................... 214
ANNEXE 9 – BREVET POUR LA MÉTHODE DE MESURE SIMULTANÉE DE LA
FRACTION DE SOLIDES ET DE LA COMPOSITION CHIMIQUE DE LA PHASE
GAZEUSE ......................................................................................................................... 257
xvi
LISTE DES FIGURES
Figure 2-1: Réacteur à lit fluidisé .................................................................................................. 17
Figure 2-2: Mesure simultanée des particules et de la composition chimique du gaz .................. 18
Figure 2-3: Spectromètre FT-IR .................................................................................................... 22
Figure 2-4: Appareil expérimental avec sonde à fibres-optiques infrarouges .............................. 23
Figure 2-5: Réacteur à lit fluidisé .................................................................................................. 24
Figure 4-1: Fluidized bed reactor .................................................................................................. 30
Figure 4-2: Conversion and YCO2 axial profiles at 923 K and 973 K ........................................... 33
Figure 4-3: XCH4 axial profiles at various fluidized bed temperatures .......................................... 34
Figure 4-4: Comparison between the model and experiments at TB = 973, 1023,1073
and 1123 K ............................................................................................................................ 35
Figure 4-5: Conversion in the injector region ............................................................................... 40
Figure 5-1: YC3H8 and GSU axial profiles in the freeboard region ................................................. 55
Figure 5-2: Axial temperature profile inside the fluidized bed and freeboard regions ................. 56
Figure 5-3: Experimental and predicted YC3H8 axial profiles for TB = 853 K and
Ug = 0.17 m/s ......................................................................................................................... 60
Figure 5-4: Experimental and predicted YC3H8 axial profiles for TB = 923 K and
Ug = 0.17 m/s:........................................................................................................................ 61
Figure 5-5: Performance of the mechanism of Marinov et al. [28] at low TB (818 K and
830 K) and with the addition of H2O2 ................................................................................... 63
Figure 6-1: FT-IR apparatus .......................................................................................................... 77
Figure 6-2: Fibre-optic probe apparatus ........................................................................................ 78
Figure 6-3: Fluidized bed reactor .................................................................................................. 79
Figure 6-4: Simultaneous measurement of YC3H8 and (1-ε) .......................................................... 81
Figure 6-5: Effect of particle movement: ...................................................................................... 83
xvii
Figure 6-6: Probe calibration......................................................................................................... 85
Figure 6-7: Solids volume fraction measurement ......................................................................... 86
Figure 6-8: Simultaneous measurement of (1-ε) and YCH4 ............................................................ 87
Figure 6-9: Error on instantaneous YCH4 measurement ................................................................. 88
Figure 6-10: Sources of error on YCH4 measurement ..................................................................... 90
Figure 6-11: Effect of (1-ε) and YCH4 on the variance of YCH4 ...................................................... 91
Figure 6-12: Simultaneous measurement of YCH4 and (1-ε) in the fluidized bed .......................... 92
Figure 6-13: Absorbance spectra in fluidized bed ........................................................................ 93
xviii
LISTE DES SIGLES ET ABRÉVIATIONS
CHAPITRE IV
A Cross-sectional area (m2)
bi Parameter in equation (7) for ith species (m)
Dj Injector nozzle diameter (m)
EA Activation energy (J)
g Gravitational constant (9.81 m/s2)
HB Fluidized bed height (m)
k0 First-order reaction rate constant (m-1)
Lj Jet penetration length (m)
Qn Jet volumetric flow rate at 0oC & 101.3 kPa (litres)
R Universal gas contant (8.314 J/Kmol)
r Radial position (m)
tj Average gas residence time in injector region (s)
T Temperature (K)
TB Fluidized bed temperature (K)
T1 Lower critical temperature (K)
T2 Upper critical temperature (K)
Uj Gas injection velocity at the sparger tip (m/s)
Umf Minimum fluidization velocity (m/s)
Ug Superficial gas velocity (m/s)
v Volumetric flow rate (m3/s)
w Nozzle thickness (cm)
Xi Conversion of ith species (%)
xix
Yi Volume fraction of ith species (vol%)
Z Axial position above the distributor (m)
Zj Axial position above the injector tip (m)
Greek letters
ε Average void fraction in fluidized bed
ρb Fluidized bed density (kg/m3)
ρg Fluidizing gas density (kg/m3)
ρj Gas density at injector nozzle (kg/m3)
ρp Particle density (kg/m3)
τ Induction time (s)
CHAPITRE V
dP average particle size (µm)
GSU solids upward flux or entrainment flux (kg/m2s)
r radial position (m)
R reaction radius (0.2 m)
τB average gas residence time in fluidized bed (s)
T temperature (K)
TB fluidized bed temperature (K)
Ug superficial gas velocity (m/s)
Umf minimum fluidization velocity (m/s)
Yi volume fraction of species i (vol%)
Z axial position (m)
xx
ZF axial position in the freeboard (m)
CHAPITRE VI
A Absorbance
dp Average particle size (µm)
f Modulation frequency (kHz)
t Distance between probe tip and mirror (mm)
Yi Molar fraction of specie i (vol%)
Greek symbols
(1−ε) Solids volume fraction
λ Wavelength (µm)
σ Variance
1
INTRODUCTION
Plusieurs procédés catalytiques industriels gaz sont caractérisés par une compétition entre des
réactions homogènes en phase gazeuse et des réactions hétérogènes à la surface de particules
solides. Les réactions homogènes peuvent mener à une baisse de production en causant la
dégradation du catalyseur, des réactifs et des produits désirés. Par ailleurs, ces réactions peuvent
représenter un risque pour la sécurité lorsqu’elles sont rapides et exothermiques comme c’est le
cas pour les réactions d’oxydations. Tandis que les procédés catalytiques impliquant des réactions
d’oxydation sélectives d’hydrocarbures sont de plus en plus répandus dans l’industrie, de
nouveaux procédés sont présentement en développement. Par exemple, des recherches sur les
procédés d’oxydation d’alcanes produisant différents produits chimiques de valeur visent à
réduire l’utilisation d’alcènes, dans ces procédés, afin de diminuer les coûts. Par ailleurs, il existe
plusieurs autres types de procédés gaz/solides impliquant des réactions d’oxydation où les
réactions homogènes peuvent avoir des effets néfastes: regénération de catalyseur, gazéification,
calcination, combustion, etc. Or, les réacteurs à lit fluidisé sont utilisés ou considérés pour
plusieurs de ces procédés en raison de leur capacité à dissiper la chaleur produite localement. En
effet, le mouvement des solides minimise la création de zones locales chaudes (« hot spots »)
étant donné que la capacité thermique des solides est significativement plus élevée que celle du
gaz.
Afin de maximiser le rendement d’un procédé, les réacteurs à lit fluidisé peuvent opérer à
l’intérieur des limites d’inflammabilité en injectant l’oxydant et les hydrocarbures séparément
dans le lit fluidisé de solides. La fraction de solides élevée dans le lit, augmentant le taux de
recombinaison des radicaux libres, évite l’initiation de réactions homogènes ou réduit leur vitesse
de réaction. Toutefois, il demeure des régions où la fraction de solides est significativement plus
faible. Dans ces régions de faible fraction de solides, des réactions homogènes peuvent être
produites par de fortes concentrations de réactifs combinés à des températures et pressions
élevées (caractéristiques de nombreux procédés incluant les procédés d’oxydation). Aussi, le taux
de recombinaison des radicaux libres peut être insuffisant et ainsi empêcher l’initiation de
réactions en phases gazeuses. Dans un réacteur à lit fluidisé, les régions de faible fraction de
solides incluent principalement:
2
• La zone du distributeur et des injecteurs dans le lit fluidisé
• La phase bulle dans le lit fluidisé
• Les régions en aval du lit fluidisé (zone de désengagement, cyclone, etc.)
À proximité du distributeur et des injecteurs, l’injection de gaz produit des zones de forte
porosité. Or, lorsque les réactions homogènes en phase gazeuse sont rapides, la conversion de
réactifs dans ces régions peut être élevée. Près des injecteurs, des réactions en phases gazeuse
peuvent dégrader les réactifs, les produits désirés et produire une zone de haute température qui
désactive le catalyseur. La phase bulle est une autre région propice aux réactions homogènes,
mais elle peut également causer un court-circuit de réactifs qui sont entraînés dans la zone de
désengagement.
Des réactifs peuvent également être entraînés dans la zone de désengagement en raison de
contraintes opératoires qui limitent la conversion dans la région du lit fluidisé. Par exemple, la
sélectivité des procédés d’oxydation partielle décroît généralement avec une augmentation de la
conversion en raison de réactions secondaires qui dégradent le(s) produit(s) désiré(s). Les réactifs
restant sont emportés dans la zone de désengagement propice à l’initiation de réactions
homogènes: porosité, température et pression élevées. Dans cette région, des réactions
homogènes rapides et exothermiques peuvent mener à une explosion. La combustion dans la zone
de désengagement représente un danger pour tous les procédés gaz/solides impliquant des
réactions d’oxydation.
Il est donc primordial de minimiser l’occurrence de réactions homogènes dans les régions de
faibles fractions de solides. Toutefois, la caractérisation de ces régions et des réactions qui s’y
produisent est une tâche ardue en raison du couplage entre la fraction de solides et la composition
chimique de la phase gazeuse. Or, les méthodes de mesures actuelles évaluent ces deux
paramètres de manière indépendante et non simultanée.
3
Une technique de mesure permettant la mesure simultanée et in situ de la composition chimique
et la fraction de solides présenterait plusieurs avantages pour la conception, la caractérisation et le
dépannage de procédés. Principalement, elle permettrait de déterminer la composition dans la
phase bulle et émulsion. À proximité du distributeur et des injecteurs, les zones de faibles
fractions de solides pourrait être cartographiées et l’effet sur les réactions homogènes/hétérogènes
mesuré. La caractérisation de la zone de désengagement serait simplifiée puisque la relation entre
la porosité et la composition chimique serait saisie. Ce type de mesures ne peut être effectué que
difficilement à l’aide des techniques existantes.
L’objectif principal de cette thèse est de développer des outils de modélisation et de
caractérisation des phénomènes de réactions homogènes en phases gazeuses dans les réacteurs à
lit fluidisé.
4
CHAPITRE 1 REVUE DE LITTÉRATURE
Ce chapitre présente une revue de la littérature scientifique sur la combustion en lit fluidisé, la
combustion dans la zone de désengagement et la caractérisation des systèmes multiphasiques.
1.1 COMBUSTION EN LIT FLUIDISÉ
La combustion du méthane et du propane dans un lit fluidisé de solides inertes a fait l’objet de
plusieurs études en mode d’injection séparée (Hesketh et Davidson 1991, Van der Vaart 1988,
Van der Vaart 1988a, Friedman et Li 2005, Dounit et al. 2008, Pre et al. 1998, Zukowski 2003,
Baron et al. 2002, Dennis et al. 1982, Srinivasan et al. 1998) et en mode pré-mélangé
(Mabrouk et al. 2008, Ross et al. 2001, Sotudeh-Gharebaagh 1998, Singh et al. 1975, Stubington
et Davidson 1981, Sotudeh-Gharebaagh et Chaouki 2007). La température du lit fluidisé a été
décrite comme un paramètre important qui détermine le taux de conversion dans la région du lit.
Accroître la température entraîne un autoallumage plus rapide dans le lit fluidisé de solides. En
mode d’injection pré-mélangé, une augmentation de la température du lit fluidisé cause le
déplacement du front de combustion vers le distributeur: le front de combustion se déplace
initialement de la zone de désengagement vers les bulles qui éclatent à la surface du lit, ensuite
dans les bulles à l’intérieur du lit et finalement dans de petites bulles au niveau du distributeur
(Zukowski 2003, Dennis et al. 1982). Alors que certaines études affirment que la combustion se
produit essentiellement dans la phase émulsion (Dennis et al. 1982), la combustion du méthane et
du propane a été observée dans des lits fixes de sables à des températures supérieures à 1023 K et
973 K, respectivement (Hesketh et Davidson 1991, Sotudeh-Gharebagh et Chaouki 2003).
Deux températures critiques ont été définies pour la combustion dans un lit fluidisé de particules
inertes. En-dessous de la température critique inférieure (T1), la combustion se produit en aval de
la surface du lit fluidisé (dans la région de désengagement). Lorsque la température du lit se
trouve entre T1 et la température critique supérieure (T2), le front de combustion se déplace dans
la phase bulle à l’intérieur du lit. Finalement, lorsque la température du lit est supérieure à T2, la
combustion se produit entièrement dans le lit fluidisé de solides et à proximité du distributeur.
Pour la combustion du méthane dans un lit fluidisé de sable, les valeurs de T1 et T2 mesurées
oscillent entre 823-1018 K et 1133-1250 K, respectivement (Laviolette et al. 2010). Pour la
5
combustion du propane, T2 a été mesuré dans un lit fluidisé de sable peu profond (HB ≈ 0.1 m) en
tant que 1123 K (Van der Vaart 1988, Dennis et al. 1982). Toutefois, peu d’études se sont
attardées à la combustion d’autres alcanes en lit fluidisé de particules inertes.
Le taux de conversion lors de la combustion en lit fluidisé peut également être estimé à l’aide
d’un modèle qui combine l’hydrodynamique gaz/solide et la cinétique de réaction. Plusieurs
modèles d’hydrodynamique de la phase gazeuse ont été proposés dans la littérature et ses
modèles varient en complexité (Laviolette et al. 2010). L’hydrodynamique de la phase gazeuse
varie entre le réacteur piston et un réacteur parfaitement mélangé selon les conditions d’opération
et la nature de la réaction. Une étude récente de traceur gazeux par Lorences et al. (2006) a
démontré que l’hydrodynamique de la phase gazeuse est comparable à celle d’un réacteur piston
lorsque Ug ≤ 6 Umf.
La région à proximité du distributeur et des injecteurs est caractérisée par un transfert de masse
plus intense comparativement au lit fluidisé (Laviolette et al. 2010). Lorsque les réactions sont
rapides, comme la combustion, une conversion importante peut survenir à l’extrémité des
injecteurs. En conséquence, la modélisation de la réaction près des injecteurs et dans le lit fluidisé
nécessite deux modèles hydrodynamiques distincts. Pour un injecteur orienté vers le bas, les
études portant sur l’auto-allumage d’hydrocarbures dans des écoulements à contre-courant
mènent à une meilleure compréhension des paramètres pouvant influencer la conversion à
l’extrémité de l’injecteur (Humer et al. 2002, Fotache et al. 1999, Jomaas et al. 2005). Ces études
démontrent que l’inflammabilité et la conversion décroissent avec la diminution de la
température et l’augmentation de la vitesse des gaz à contre-courant. En effet, les réactions
menant à l’auto-allumage des hydrocarbures impliquent des espèces chimiques intermédiaires
très réactives. Or, leur durée de vie peut être du même ordre que l’échelle de temps
caractéristique du transfert de masse et d’énergie. Afin de modéliser la région de l’injecteur, des
modèles à une et deux phases ont été proposés dans la littérature (Laviolette et al. 2010).
Plusieurs types de modèles de cinétique de réaction sont disponibles dans la littérature pour les
systèmes homogènes en phase gazeuse: modèles de microcinétique et modèles globaux
6
(Dounit et al. 2008, Pre et al. 1998, Ross et al. 2005). Certains mécanismes de microcinétique ont
été modifiés afin d’inclure les réactions de recombinaison des radicaux libres à la surface des
particules solides (Sotudeh-Gharebagh et Chaouki 2003, Jeng et al. 2001). Les modèle de
microcinétique sont complexes, mais essentiel afin de prédire la production et la consommation
d’espèces chimiques intermédiaires. Toutefois, dans des conditions de combustion pauvre, la
conversion de l’hydrocarbure dans un lit fluidisé est bien représentée par une cinétique globale de
premier ordre (Sotudeh-Gharebagh et Chaouki 2003, Foka et al. 1994, Chaouki et al. 1999). Des
expressions de cinétiques globales pour une combustion homogènes en phase gazeuse ont été
proposées pour le méthane, mais demeurent rares pour l’éthane et les alcanes plus complexes
(C2+) (Ross et al. 2004).
Il peut être possible d’exprimer la cinétique globale de premier ordre ou la conversion pour les
alcanes C2+ à l’aide d’une seule expression ou corrélation. Plusieurs études ont fait état de
similitudes entre la chimie de combustion des alcanes : le taux de combustion est contrôlé par les
mêmes réactions radicalaires. Ces réactions peuvent être résumées par les réactions élémentaires
suivantes à températures intermédiaires (~ 850 K à 1200 K) (Horning et al. 2002,
Burcat et al. 1971, Westbrook 2000):
MHOMOH +→++ 22 (1.1)
222 OHRHORH +→+ (1.2)
MOHOHMOH ++→+22 (1.3)
Dans les réactions ci-dessus, le symbole RH correspond à un alcane, R est un alkyle et M est
partenaire de collision. Saxena et al. (2007) ont démontrés que le temps d’induction à haute
température (> 1200 K) du propane et alcanes plus complexes peut être estimé de manière précise
en employant les paramètres décrivant la vitesse des réactions (2) à (4):
22222 OOHHOHO +→+ (1.4)
7
Plusieurs études ont démontré que le temps d’induction des alcanes C2-C5 mesuré dans des tubes
à chocs exhibe une dépendance similaire à la température, pression and composition chimique:
une corrélation avec une énergie d’activation constante suffit (Horning et al. 2002,
Burcat et al. 1971). Des études précédentes rapportent également qu’à une stoichiométrie
température et pression similaire, le temps d’induction décroît telle que C2H6 < n-C4H10 < C3H8
(Horning et al. 2002, Burcat et al. 1971, Lamoureux et al. 2002). D’un autre côté, le méthane est
caractérisé par des temps d’induction significativement supérieurs résultant de la production d’un
radical méthyle dans l’équation (1.2). Le radical méthyle se recombine pour former de l’éthane
dans une réaction de terminaison. Si la chimie de combustion et d’autoallumage des alcanes C2+
démontre de telles similarités, un parallèle peut possiblement être fait entre leurs cinétiques de
réaction respective.
1.2 COMBUSTION DANS LA ZONE DE DÉSENGAGEMENT
Les procédés en lit fluidisé sont souvent opérés à l’intérieur des limites d’inflammabilité afin
d’augmenter le rendement. Dans ce cas, l’oxydant et l’hydrocarbure sont injectés séparément à
l’intérieur du lit fluidisé de solides. La fraction de solides élevée à l’intérieur du lit inhibe les
réactions homogènes en phase gazeuse (combustion) tout en promouvant les réactions sélectives
en phase hétérogène. Toutefois, en raison de contraintes opératoires, les procédés peuvent être
opérés avec une conversion partielle des réactifs à l’intérieur du lit de solides. Par exemple, les
réactions d’oxydation sélective sont caractérisées par une sélectivité qui décroît en fonction de
l’augmentation de la conversion (Botella et al. 2002, Lorences et al. 2006a, Malleswara Rao and
Deo 2007, Hutchenson et al. 2009). Par conséquent, en aval du lit fluidisé – dans la zone de
désengagement, les cyclones et les multiples conduites associées – la mixture gazeuse est
potentiellement explosive puisque caractérisée par une concentration de réactifs (hydrocarbure et
oxygène), une température, une pression et une porosité élevées (Van der Vaart 1988,
Van der Vaart 1988a). Minimiser le risque de réactions homogènes dans ses régions constitue un
enjeu important.
Plusieurs indices d’explosions sont disponibles dans la littérature pour des systèmes homogènes
en phase gazeuse: limites d’inflammabilité, température d’auto-allumage et corrélations de temps
d’induction (Brokaw et Jackson 1955, Chang et al. 1959, Burcat et Lifshitz 1971,
8
Burcat et al. 1971, Cathonnet et al. 1981, Kin et Soo Shin 2001, Penyazkov et al. 2005,
Horning et al. 2002, Zhukov et al. 2005, Freeman et Lefebvre 1984, Cadman et al. 2000).
Hesketh et al. (1991) ont démontré que les corrélations de temps d’induction peuvent prédire
avec précision les délais d’auto-allumage au-dessus d’un lit fixe pour TB ≤ 973 K et lorsque la
combustion dans le lit est négligeable. Toutefois, utiliser les résultats de ces corrélations afin de
prédire le temps d’induction dans des réacteurs à lit fluidisé est incertain. En effet, les particules
sont emportées en aval du lit fluidisé où elles inhibent les réactions homogènes en phase gazeuse.
De plus, la distribution axiale et radiale des solides au-dessus du lit n’est pas uniforme: la
quantité de solides emportés décroît exponentiellement à partir de la surface du lit fluidisé et
atteint une valeur constante à une distance appelée « hauteur limite de désengagement des
particules » (Wen et Chen 1982, Kunii et Levenspiel 1990). La zone de désengagement peut
également être caractérisée par des gradients radiaux et axiaux de température et de
concentrations d’espèces chimiques. Finalement, la conversion partielle des réactifs dans le lit
fluidisé de solides produits des espèces chimiques intermédiaires et des radicaux libres qui
peuvent participer aux réactions dans la zone de désengagement.
Afin de prédire le temps d’induction et la vitesse de combustion d’une mixture inflammable dans
la zone de désengagement et ce pour différentes conditions opératoires, un modèle de combustion
combinant l’hydrodynamique gaz/solide, la cinétique de réaction, le gradient de température et la
condition limite adéquate doit être développé. Des modèles en régime permanent ont été proposés
par De Lasa et al. (1979), Chen et al. (1982), Walsh et al. (1982), Sotudeh-Gharebagh et
Mostoufi (2003), Dounit et al. (2008), and Hartman et al. (2010) pour représenter différentes
réactions dans la zone de désengagement. Toutefois, ces modèles utilisent des cinétiques globales
qui sont incapables de simuler la chimie radicalaire et le temps d’induction. Également, ces
modèles utilisent des corrélations afin d’estimer la fraction de solides, ce qui risque d’introduire
des erreurs significatives.
Un mécanisme de microcinétique est nécessaire afin de caractériser le temps d’induction et de
considérer l’effet des espèces chimiques intermédiaires sur les réactions. Plusieurs modèles de
microcinétique pour des systèmes homogènes en phase gazeuse sont disponibles dans la
9
littérature pour les régimes de basse température (T < 600 K) (Buda et al. 2006), de température
intermédiaire (650-1000K) (Marinov et al. 1998, Dagaut et al. 1987) et de température élevée
(T > 1000 K) (Qin et al. 2000, Sung et al. 1998, San Diego Mechanism, Smith et al. 2010). Ainsi,
la cinétique de combustion du propane subit une transition d’un mécanisme à faible température
(T < 600 K) à un mécanisme à température intermédiaire (T > 650 K). Ces deux régimes sont
séparés par une région à coefficient de température négatif (600 – 650 K). La production de
flammes froides a été observée à température faible et intermédiaire. La création de flammes
froides est associée à un changement d’agent de réaction radicalaire en chaîne des
hydroperoxydes au peroxyde d’hydrogène (Ranzi et al. 1994, Wilk et al. 1986). Certains
mécanismes de microcinétique ont été modifiés afin d’inclure les réactions de recombinaison des
radicaux libres (Jeng et al. 2001, Sotudeh-Gharebagh et Chaouki 2003).
Un modèle de microcinétique a été utilisé par Hutchenson et al. (2010) afin de prédire la
conversion d’oxygène dans la zone de désengagement d’un réacteur à lit fluidisé pilote (0.128 m
I.D.) contenant un catalyseur de pyrophosphate de vanadium (VPP) pendant l’oxydation sélective
de n-butane. Les conversions calculées ont été comparées aux résultats expérimentaux: les
prédictions étaient significativement plus faibles que les valeurs mesurées. Deux explications ont
été avancées par les auteurs:
1. Des réactions importantes en phase homogène et hétérogènes étaient manquantes dans le
modèle de microcinétique
2. La production de radicaux libres dans la région du lit fluidisé de solides n’était pas
considérée dans le modèle.
Toutefois, cette étude était basée sur une caractérisation limitée de la zone de désengagement: la
température et la composition chimique de la phase gazeuse étaient mesurées uniquement à une
position de telle sorte que la zone de désengagement était considérée parfaitement mélangée.
Le modèle de réaction doit être validé à l’aide de données expérimentales où les paramètres
influençant la combustion dans la zone de désengagement sont mesurés précisément:
10
composition chimique de la phase gazeuse, la fraction de solides et la température. De plus,
puisque ces paramètres sont dépendants l’un de l’autre, ils doivent être mesurées simultanément,
ce qui représente un défi important.
1.3 CARACTÉRISATION DES SYSTÈMES MULTIPHASIQUES
La mesure simultanée de la fraction de solides et de la composition chimique de la phase gazeuse
est importante afin de caractériser les lits fluidisés gaz/solides. Or elle l’est également pour
plusieurs autres types de systèmes. Les procédés multiphasiques sont répandus dans plusieurs
industries: pétrochimique, chimique, pharmaceutique, alimentaire, pâte et papier, etc. Ils sont
également présents dans l’ensemble de la chaîne de production: génération d’énergie (chambres à
combustion), les procédés de transformation (gazéifieurs, réacteurs chimiques, mélangeurs et
sécheurs), les procédés de purification/séparation (cyclones et précipitateurs), l’évacuation de
déchets (cheminés), etc. La caractérisation des ces systèmes par la mesure de la composition
chimique et la fraction de solides est critique à des fins de contrôle, d’optimisation et de
dépannage dans les laboratoires et les industries. Ces mesures sont également essentielles pour le
développement de modèles hydrodynamiques (gaz/solide/liquide) et de cinétique de réaction.
Présentement, la composition et la fraction de solides sont mesurées par deux techniques
distinctes.
Plusieurs techniques invasives et non-invasives de mesures ont été développées afin de mesurer
la fraction de solides dans les systèmes di- et tri-phasiques. Les techniques invasives nécessitent
l’introduction d’une sonde pour mesurer localement la capacité (Werther et Molerus 1972, Yates
et Simons 1994, Hage et Werther 1997), la conductivité (Begovich et Watson 1978, Uribe-Salas
et Gomez 1994, Liu et al. 2007), la fluorescence (Huizenga et al. 1998) ou la diffusion d’un rayon
lumineux dans le milieu (Yates et Simons 1994, Cutolo et al. 1990, Ishida et Tanaka 1982,
Hu et al. 1986, Esmaeili et al. 2008). La fraction de solides est dérivée de la mesure qui peut être
effectuée in situ et en temps réel. Dans les systèmes liquide/solide ou gaz/liquide/solide, la phase
coulis peut être également échantillonnée par l’entremise de sondes spéciales afin de mesurer la
fraction volumique de liquide et solide (Huizenga et al. 1998, Gandhi et al. 1999). Les techniques
non-invasives incluent la tomographie (Chaouki et al. 1997, Pugsley et al. 2003,
Marashdeh et al. 2008, Razzak et al. 2009), la densitométrie (Esmaeili et al. 2008) et l’analyse de
11
la distorsion d’ondes ultrasoniques (Uchida et al. 1989). Ces techniques produisent des cartes de
fraction de solides moyennées dans le temps ou l’espace avec une résolution variable.
La mesure de la composition chimique dans les phases gazeuse, liquide et solide peut être
effectuée en continu ou de manière intermittente (Gandhi et al. 1999) à l’aide d’instruments
analytiques (chromatographe, spectromètre, etc) et peut nécessiter l’arrêt du procédé
(Luypaert et al. 2007, Swarbrick 2007). L’échantillonnage peut présenter plusieurs désavantages:
perte de production suite à un arrêt, perturbation du système en raison de l’introduction de la
sonde (l’erreur de mesure peut être de l’ordre de 5% à 100%), destruction de l’échantillon et perte
du produit (Luypaert et al. 2007, Swarbrick 2007). Dans le contexte du « process analytical
technology » (PAT) de la « U.S. Food and Drug Administration » (FDA), plusieurs publications
présentent le développement d’application en spectroscopie infrarouge permettant la mesure de la
composition chimique dans les systèmes multiphasiques (Armaroli et al. 2004,
Kondo et al. 2007). Des techniques similaires ont également été mises au point afin de
caractériser les réactions se produisant à la surface de catalyseurs solides (Corveleyn et al. 1997,
Rantanen et al. 2000a). Ces applications fonctionnent en mode transmission ou réflexion diffuse
(à la surface des solides) et peuvent mesurer la composition chimiques dans la phase dispersée
(particules solides et gouttelettes liquides) et la phase continue (liquide ou gazeuse)
(Rantanen et al. 2000, Cogdill et al. 2005, Triadaphillou et al. 2007, Andersson et al. 1999,
Kirsch et Drennen 1995). La spectroscopie infrarouge présente l’avantage d’effectuer des
mesures in situ et en temps réel par l’entremise d’une sonde à fibres optiques.
Une technique de mesure permettant la mesure simultanée de la composition chimique et la
fraction de solides présenterait de nombreux avantages. Premièrement, la mesure de ces deux
paramètres nécessiterait un seul instrument comparativement à deux, ce qui réduirait
potentiellement la complexité de la mesure et le coût des équipements. Deuxièmement, une telle
méthode offrirait beaucoup plus d’information sur le procédé puisque la fraction de solides et la
composition chimique sont dépendantes. En effet, ceux deux paramètres sont liés par la cinétique
de réaction et l’hydrodynamique. Les particules solides influencent les réactions au moyen
d’effets inhibitifs et/ou catalytiques et également par le biais du bilan calorifique. De plus, au
12
niveau de l’hydrodynamique, les systèmes multiphasiques sont souvent composés de différentes
régions caractérisées par une fraction de solides distinctes. La mesure de la composition chimique
dans ses régions est fondamentale à la description d’un procédé. Par exemple, les lits fluidisés
gaz/solide à bulles et turbulents possèdent une phase dense (émulsion) et une phase diluée (bulle).
Les bulles causent un court-circuit des réactifs qui réduit le rendement d’un procédé. La mesure
simultanée de la fraction de solide et la composition chimique dans la phase gazeuse
permettraient de déterminer la composition dans la phase dense et diluée. Ce type de mesure peut
difficilement être effectué à l’aide des techniques existantes.
La spectroscopie infrarouge présente un potentiel important pour la mesure simultanée de
fraction de solides et de la composition chimique de la phase gazeuse. Les applications de
spectroscopie infrarouge actuelles sont affectées par la phase solide. La fraction de solides et la
taille des particules affectent l’interaction entre la lumière et le système en influençant sur la
longueur du trajet d’absorption et la taille d’échantillon (Corveleyn et al. 1997,
Cogdill et al. 2005). En effet, une augmentation de la taille des particules diminue l’intensité de
lumière diffuse et accroît le trajet d’absorption du rayon lumineux produisant une augmentation
de l’absorbance et un déplacement de la ligne de base en mode réflexion diffuse
(Cogdill et al. 2005). Conséquemment, les techniques actuelles de spectroscopie infrarouge
utilisées dans les systèmes multiphasiques nécessitent que l’échantillon soit uniforme dans le
temps (d’une mesure à l’autre). Ceci est réalisé concrètement en effectuant des mesures dans des
systèmes immobiles comme les comprimés pharmaceutiques (Andersson et al. 1999, Kirsch et
Drennen 1996, Abrahamsson et al. 2005, Hailey et al. 1996, Sonja Sekulic et al. 1998). Pour les
systèmes multiphasiques où la phase solide est en mouvement (lits fluidisés et mélangeurs, par
exemple), les mesures nécessitent un arrêt du procédé (Sonja Sekulic et al. 1996,
Frake et al. 1997), la synchronisation de l’acquisition du spectre avec une position spécifique de
la cuve (Bellamy et al. 2008) ou la mesure dans les régions de faible porosité (Cui et al. 2001).
Des mesures de porosité ont également été effectuées dans des mélangeurs symétriques afin
d’assurer l’uniformité de la fraction de solides une fois le système parfaitement mélangé
(De Paepe et al. 1997). Finalement, le traitement mathématique des spectres mesurés est souvent
utilisé afin de minimiser les effets du déplacement de la ligne de base causé par le mouvement
des solides (Cogdill et al. 2005). Toutefois, les systèmes multiphasiques sont généralement
13
caractérisés par des fractions de solides qui varient dans le temps et l’espace. Par conséquent, la
mesure in situ et en temps réel de la composition chimique dans les différentes régions du
système nécessite la détermination simultanée de la fraction de solides. Le taille d’échantillon
peut être indépendant des propriétés des particules si les fibres optiques émettrices et réceptrices
sont inclinées l’une vers l’autre (Berntsson et al. 2001). Ce principe peut être utilisé sur une fibre
optique infrarouge pour mesurer la composition chimique.
Dans les systèmes multiphasiques, le mouvement des particules solides rend les mesures
spectroscopiques et leurs interprétations plus complexes. La mesure d’absorbance est influencée
par les variations de fractions de solides et la distance moyenne entre l’échantillon et la sonde.
Lorsque les échantillons sont en mouvement, les régions du spectre peuvent être influencées
différemment par les hétérogénéités locales. L’effet du mouvement de particules sur la
spectroscopie par transformée de Fourier a été étudié par De Paepe et al. (1997),
Berntsson et al. (2001) and Andersson et al. (2005). Il a été démontré que le mouvement des
solides produit du bruit dans le spectre mesuré à certains nombres d’ondes qui dépendent de
l’échelle de temps du balayage spectral comparativement à la vitesse des particules. Le nombre
d’onde auquel le bruit apparaît augmente en fonction de l’accroissement de la fréquence de
modulation. Ainsi, l’effet des solides peut être complètement éliminé aux nombres d’ondes
d’intérêts si la fréquence de modulation est suffisamment élevée.
14
CHAPITRE 2 PRÉSENTATION DES ÉTAPES DE TRAVAIL
Minimiser l’occurrence de réactions homogènes dans les régions de faibles fractions de solides
nécessite une bonne capacité de modélisation des réactions dans le lit fluidisé afin d’y obtenir une
conversion élevée et empêcher la formation d’une mixture inflammable dans la zone de
désengagement.
Si des contraintes opératoires limitent la conversion qui peut être obtenue dans la région du lit
fluidisé, il est essentiel de caractériser le risque d’auto-allumage dans la zone de désengagement
en fonction des conditions qui y règnent: composition chimique de la phase gazeuse, profils de
fraction de solides, température et pression.
La fraction de solides et la composition chimique de la phase gazeuse sont deux paramètres
dépendants qui doivent être mesurés simultanément. Une telle mesure est critique pour la
caractérisation de la zone de désengagement et des régions de faibles fractions de solides dans le
lit fluidisé: les régions à proximité des injecteurs et la phase bulle. Or, une telle mesure est
difficile avec les techniques actuellement disponibles.
Les objectifs spécifiques de cette thèse sont:
1. Étudier expérimentalement la combustion d’hydrocarbures dans un lit fluidisé de solides
inertes afin d’identifer les paramètres de fluidization menant à une conversion complète et
d’améliorer les moyens disponibles pour y modéliser les réactions.
2. Étudier l’auto-allumage de mixtures hydrocarbure/air dans la zone de désengagement et
tenter de modéliser les réactions avec une microcinétique afin de reproduire le temps
d’induction et le taux de réaction.
3. Développer une nouvelle méthode pour mesurer simultanément la fraction de solides et la
composition chimique de la phase gazeuse. Cette méthode devrait permettre d’effectuer
des mesures en ligne et in situ dans des systèmes multiphasiques.
15
Afin de répondre à ces objectifs, le chapitre 4 décrit une étude expérimentale sur la combustion
des alcanes C1-C4 dans un lit fluidisé de particules inertes. Les températures nécessaires à la
conversion complète des alcanes C1-C4 dans la région du lit fluidisé (températures critiques de
transition) sont mesurées. De plus, une cinétique de réaction globale et de premier ordre est
mesurée pour chacun des alcanes. Finalement, un modèle pour prédire la conversion dans la
région de l’injecteur et la région du lit fluidisé est proposé.
Le chapitre 5 présente une étude sur la combustion du propane dans la zone désengagement d’un
lit fluidisé. Le temps d’induction et le taux de réaction du propane sont modélisés avec plusieurs
modèles de microcinétique de combustion homogène en phase gazeuse.
Le chapitre 6 relate le développement d’une nouvelle méthode de spectroscopie infrarouge qui
permet la mesure simultanée de la fraction de solides et de la composition chimique dans la phase
gazeuse. La méthode est utilisée avec une sonde à fibres optiques infrarouges afin d’effectuer des
mesures in situ et en ligne dans des systèmes gaz/solides. Des mesures sont effectuées dans un lit
fluidisé afin de mesurer la fraction volumétrique d’un traceur gazeux dans la phase bulle et la
phase émulsion.
Le chapitre 3 présente la méthodologie pour effectuer ces travaux.
16
CHAPITRE 3 MÉTHODOLOGIE
Ce chapitre présente la méthodologie suivie pour effectuer les expériences et la modélisation.
3.1 EXPÉRIENCES DE COMBUSTION EN LIT FLUIDISÉ
Les expériences sur la combustion en lit fluidisé des alcanes C1-C4 ont été effectuées sur le
réacteur à lit fluidisé de la Figure 2-1. La région du lit fluidisé de solides avait un diamètre
interne de 0.2 m et une hauteur d’environ 1.0 m. Des particules inertes de sables de silice
(Geldart B, ρs = 2650 kg/m3, dp = 290 µm et Umf = 0.08 m/s) ont été utilisées dans le lit. Le lit de
particules était fluidisé avec de l’air injecté par le distributeur et l’hydrocarbure (méthane, éthane,
propane ou n-butane) était acheminé séparément par un injecteur orienté vers le bas (4.1 mm
D.I.). L’extrémité de cet injecteur se trouvait à une distance de 0.1 m au-dessus du distributeur.
Avant chaque expérience, le réacteur à lit fluidisé était chauffé à la température désirée par
l’entremise d’un brûleur de gaz naturel connecté à la boîte à vent. De l’huile lourde était
également brûlée dans le lit afin d’accélérer le processus de chauffage. Une fois la température du
lit atteinte, l’injection d’huile lourde et de gaz naturel était arrêtée et le lit fluidisé était chauffé
avec du propane pendant au moins une heure afin de brûler tous les résidus d’huile lourde.
Les expériences ont été effectuées à une vitesse superficielle de gaz de 0.4 m/s et cinq
températures de lit : 923 K, 973 K, 1023 K, 1073 K et 1123 K. La température à l’intérieur du
réacteur était mesurée à neuf positions axiales différentes à l’aide de thermocouples. Le gaz était
échantillonné à six positions axiales où une mesure était effectuée à cinq positions radiales. Le
gaz échantillonné était ensuite analysé par un analyseur Ultramat™ de Siemens (CO/CO2) et un
micro-chromatographe en phase gazeuse CP-4900™ de Varian afin de mesurer la fraction
volumétrique des espèces chimiques suivantes: H2, CH4, C2H4, C2H6, C3H8, n-C4H10, O2, N2, CO
et CO2. Pour chaque position axiale, une moyenne de la composition chimique était calculée sur
la surface.
Du gaz était également échantillonné dans la section de désengagement afin de mesurer la
fraction volumétrique d’oxygène à l’aide d’un Oxymat™ de Siemens. Lors des expériences, le
débit de la mixture d’hydrocarbure était ajusté afin d’obtenir un débit calorifique constant de
17
3.8 kW (calculé à partir du pouvoir calorifique inférieur). Pour le méthane, l’éthane, le propane et
le n-butane, les fractions volumétriques correspondantes à l’extrémité de l’injecteur étaient
respectivement de 5.1%, 2.5%, 1.9% et 1.3%. Ceci correspondait à des vitesses d’injection de
52 m/s, 25 m/s, 18 m/s et 13 m/s pour le méthane, l’éthane, le propane et le n-butane,
respectivement.
Figure 3-1: Réacteur à lit fluidisé
3.2 EXPÉRIENCES DE COMBUSTION DANS LA ZONE DE
DÉSENGAGEMENT
Les expériences sur la combustion du propane dans la zone de désengagement ont été effectuées
dans le même réacteur à lit fluidisé (figure 3-1) avec les mêmes particules inertes de sables de
silice. La zone de désengagement était caractérisée par un diamètre interne de 0.2 m. Le propane
était injecté séparément au moyen de l’injecteur orienté vers le bas. Pour toutes les expériences,
une vitesse superficielle de 0.17 m/s (Ug = 2.1 Umf) et de faibles températures de lit
3 4
Sondes de prélèvement non-isocinétique
0.48 m
0.2 m
0.97 m
0.38 m
1 1er
cyclone 2 2
ième cyclone
3 Brûleur de gaz naturel 4 Boîte à vent 5 Lit fluidisé de solides
0.6 m
0.2 m
Air + Gaz naturel
C3H8
r
5
Analyseur de O2
Analyseur CO/CO2
z
1
2
Chromatographe en phase gazeuse
18
(818 K ≤ TB ≤ 923 K) ont été utilisées. La surface du lit fluidisé de solides et la zone de
désengagement ont été caractérisées via la mesure simultanée du flux d’entraînement de
particules, de la composition chimique de la phase gazeuse et de la température.
Les particules solides et le gaz étaient échantillonnés au moyen d’une sonde de prélèvement non-
isocinétique similaire à celles décrites par Rhodes et Laussmann (1992), Reinhardt et al. (1999) et
Van der Meer et al. (2000). Le flux de solides était déduit de la masse de solide récupérée:
tA
mG S
D,SU =
De plus, la composition chimique du gaz était déterminée avec un chromatographe en phase
gazeuse CP-4900™ de Varian et un Ultramat™ de Siemens. La figure 3-2 illustre le système
d’échantillonnage. La température dans le réacteur était mesurée par 10 thermocouples
positionnés sur la longueur du réacteur. Plus d’informations sur la sonde à prélèvement non-
isocinétique et les techniques utilisées afin de mesurer simultanément le flux de solides, la
composition chimique de la phase gazeuse et la température sont disponibles dans l’annexe 2.
Figure 3-2: Mesure simultanée des particules et de la composition chimique du gaz
Récupération des particules solides
Chromatographe
et
Ultramat
Sonde de prélèvement non-isocinétique
Succion
19
3.3 MODÉLISATION DE LA COMBUSTION EN LIT FLUIDISÉ
L’hydrodynamique de la phase gazeuse dans le lit fluidisé a été représentée à l’aide d’un modèle
composé de deux réacteurs pistons en séries. De plus, une cinétique de réaction globale de
premier ordre a été utilisée. Ces choix se justifient tel que mentionné dans la revue de littérature,
car ils permettent de représenter de façon appropriée l’hydrodynamique de la phase gazeuse dans
un lit fluidisé de faible vitesse superficielle et une combustion pauvre (Lorences et al. (2006),
Laviolette et al. (2010)).
Le lit fluidisé a été séparé en deux zones distinctes:
1. La région de l’injecteur caractérisée par un taux de mélange gaz/gaz et gaz/solide élevé
2. Le lit fluidisé
Ces deux régions ont été modélisées par un réacteur piston (lit fluidisé) avec un décalage d’axe
(effet de la conversion à l’injecteur) de sorte que la conversion de l’hydrocarbure était calculée
par les équations suivantes :
BedFluidized,iInjector,ii XXX += (3.1)
( )( )ij bZk
i eX+′−
−= 1100 (3.2)
RTE
RTE AA
ekev
kAk
−−′==′ 0
0 (3.3)
Le paramètre Zj correspond à la position où le mélange entre l’air et l’hydrocarbure survient dans
le lit fluidisé et où les réactions d’oxydation débutent. Afin d’évaluer Zj pour une combustion par
des injections séparées, le jet à l’extrémité de l’injecteur a été caractérisé par la profondeur de
pénétration calculée au moyen de six corrélations différentes (Zenz 1968, Benjelloun et al. 1995,
Shen et al. 1990, Shakhova 1968, Merry 1971, Hong et al. 1997, Okhotskii 2001,
Penyazkov 2005).
20
3.4 MODÉLISATION DE LA COMBUTION DANS LA ZONE DE
DÉSENGAGEMENT
Afin de représenter la combustion dans la zone de désengagement d’un lit fluidisé, un modèle
utilisant une microcinétique de réaction en phases gazeuses a été développé. Tel que mentionné
dans la revue de littérature, la cinétique de combustion du propane subit une transition d’un
mécanisme à faible température (T < 600 K) à un mécanisme à température intermédiaire
(T > 650 K). Ces deux régimes sont séparés par une région à coefficient de température négatif
(600 – 650 K). La production de flammes froides a été observée à température faible et
intermédiaire. La création de flammes froide est associée à un changement d’agent principal de
réaction radicalaire en chaîne des hydroperoxydes au peroxyde d’hydrogène (Ranzi et al. 1994,
Wilk et al. 1986).
Les expériences effectuées lors de cette recherche utilisaient des températures intermédiaires où
la réaction radicalaire principale serait la dissociation du peroxyde d’hydrogène en deux radicaux
d’hydroxyle (Ranzi et al. 1994, Wilk et al. 1986). Afin de modéliser le cinétique de combustion,
six mécanismes de microcinétique bien établis dans la littérature scientifique ont été utilisés et
comparés: le mécanisme de Dagaut et al. (1987), Marinov et al. (1998), San Diego (2010),
GRI version 3 (2010), Sung et al. (1998) et Qin et al. (2000).
Le modèle de Dagaut et al. (1987) (50 espèces chimiques et 274 réactions élémentaires) a été
validé au moyen de résultats expérimentaux obtenus dans un réacteur auto-agité par jets gazeux
pour un large spectre de coefficients d’équivalence (0.15-4) et de températures (900-1200 K). Le
modèle de Marinov et al. (1998) (126 espèces chimiques et 638 réactions élémentaires) a
démontré une bonne concordance avec des résultats expérimentaux d’oxydation de NO et
d’hydrocarbures dans un réacteur piston atmosphérique et une plage de températures de 600 à
1100 K. Ce mécanisme inclut la cinétique de combustion de l’hydrogène, le méthane, l’éthylène,
l’éthane, le propane et l’éthanol. Le mécanisme de San Diego (2010) (46 espèces chimiques et
235 réactions élémentaires), la troisième version du mécanisme GRI (2010) (53 espèces
chimiques et 325 réactions élémentaires), le mécanisme de Sung et al. (1998) (92 espèces
21
chimiques et 621 réactions élémentaires) et le modèle de Qin et al. (2000) (70 espèces chimiques
et 463 réactions élémentaires) ont également été utilisés. Ces quatre modèles ont été comparés à
des résultats expérimentaux (temps d’induction, vitesses de flamme et gradients axiaux d’espèces
chimiques) obtenus à des températures supérieures à celles utilisées lors des expériences de cette
thèse. Toutefois, ces mécanismes contiennent les réactions élémentaires pour décrire la
combustion à températures faibles et intermédiaires.
L’hydrodynamique de la phase gazeuse dans la zone de désengagement a été modélisée au
moyen d’un réacteur piston. Les profils de température mesurés expérimentalement en régime
permanent on été utilisés dans le modèle réactionnel et les équations ont été résolues à l’aide du
logiciel ChemKin 4.1. Les mécanismes de microcinétique ne tenaient pas compte des réactions en
phase hétérogène qui auraient pu avoir lieu à la surface des particules solides.
Deux séries de simulations ont été effectuées. Dans la première série, la zone de désengagement a
été modélisée par un réacteur piston. La composition chimique de la phase gazeuse a, quant à
elle, été mesurée à la surface du lit fluidisé et a été utilisée comme condition frontière. Dans la
seconde série de simulations, un second réacteur piston a été ajouté, en amont, afin de simuler les
réactions ayant lieu dans la région du lit fluidisé de solides et la production d’agents de réaction
radicalaire. La composition chimique parfaitement mélangée à l’extrémité de l’injecteur a été
utilisée en tant que condition frontière à l’entrée du réacteur piston simulant le lit fluidisé de
solides. De plus, le temps de résidence moyen du gaz a été ajusté individuellement afin d’obtenir
la conversion adéquate dans le lit fluidisé. La mixture, à la sortie du premier réacteur piston, (lit
fluidisé) était ensuite injectée à l’entrée du second (zone de désengagement).
3.5 DÉVELOPMENT D’UNE MÉTHODE POUR LA MESURE
SIMULTANÉE DE LA FRACTION DE SOLIDES ET DE LA
COMPOSITION CHIMIQUE DE LA PHASE GAZEUSE
Des expériences ont été effectuées avec un FT-IR (modèle Excalibur 3100™ de Varian) équipé
d’un détecteur à haute sensibilité tellurure de mercure-cadmium (TMC) refroidi à l’azote liquide.
Dans une première série de tests, le FT-IR a été utilisé avec une cellule de mesure placée dans un
22
compartiment à échantillons tel qu’illustré dans la Figure 3-3. Le rayon infrarouge venant de
l’interféromètre était transmis à travers la cellule de mesure. Le spectre d’absorbance résultant
était mesuré par le détecteur TMC. Avant chaque expérience, le signal de fond était mesuré en
purgeant la cellule de mesure avec de l’air et en prenant la moyenne de 20 spectres. Ensuite, un
système gaz/solide était simulé dans la cellule de mesure en injectant une mixture de
1.8 vol% C3H8 + 98.2 vol% air et en plaçant une toile métallique dans la trajectoire du rayon
infrarouge incident (épaisseur de 1 mm et 16% de surface ouverte).
Figure 3-3: Spectromètre FT-IR
Une seconde série d’expériences a été réalisée avec une sonde à fibres-optiques pour l’infrarouge
moyen. Des mesures in-situ de composition chimique gazeuse et de fraction volumétrique de
solides ont été accomplies à l’intérieur d’un écoulement gaz/solide de CH4/N2 et de particules de
sable de silice (dp = 290 µm). Une illustration de l’appareil expérimental est présentée dans la
Figure 3-4. La sonde était composée de deux fibres-optiques de silice fluorée (une fibre émettrice
et une fibre réceptrice) avec une ouverture numérique de 0.2 et un diamètre de cœur de 600 µm.
La sonde à fibres-optiques était connectée au FT-IR au moyen d’un Fibremate™ de la compagnie
Harrick. Un miroir plan plaqué d’or était positionné perpendiculairement à une distance de 5 mm
de l’extrémité de la sonde. Le miroir réfléchissait le rayon IR incident vers la fibre-optique
réceptrice, ce qui produisait un volume de mesure entre la sonde et le miroir tel que coloré en gris
dans la figure 3-4. Les expériences ont été effectuées en premier lieu par la mesure du signal de
fond – une mixture de CH4/N2 était injectée dans le volume de mesure et un spectre de fond,
constitué de la moyenne de 20 spectres individuels, était mesuré. Ensuite, le volume de mesure
Détecteur TMC
Interféromètre du FT-IR
Toile métallique
Cellule de mesure
(C3H8/Air)
23
était purgé avec de l’air tandis que l’entonnoir était rempli de particules de sable afin d’initier un
écoulement de solides. Les spectres d’absorbance ont été enregistrés à une fréquence de 4.5 Hz
(résolution temporelle de 0.22 s) et la fraction molaire de méthane a été mesurée simultanément à
la fraction de solides. Le taux d’écoulement de particules solides a été varié en modifiant le
diamètre de la gorge de l’entonnoir. La fraction de solides correspondante à chacune des
diamètres de gorge ont ensuite été mesurée au moyen d’une sonde à fibre-optique
conventionnelle. Quatre différentes fractions de solides ont été utilisées: 0, 0.015, 0.038 et 0.097.
De plus, quatre mixtures différentes de CH4/N2 ont été injectées à partir de cylindres de gaz
calibrés. Ces mixtures étaient caractérisées par des fractions volumétriques de méthane de 0, 0.1,
3.25 et 10.1 vol% dans l’air ou l’azote.
Figure 3-4: Appareil expérimental avec sonde à fibres-optiques infrarouges
Des expériences de traceur gazeux ont également été effectuées dans le petit réacteur à lit fluidisé
(5 cm D.I.) illustré dans la figure 3-5. La sonde à fibres-optiques a été utilisée afin de mesurer la
fraction volumétrique de traceur dans les phases bulle et émulsion. Le lit de solides était composé
de particules FCC (Umf = 2.5 mm/s and Umb = 2.7 mm/s, dp = 83 µm et 14% de fines
(dp ≤ 44 µm)) et il était fluidisé avec un écoulement d’azote d’une vitesse de 2.6 mm/s. Le lit
Fibre-optique émettrice
Fibre-optique réceptrice FT
-IR
Miroir plan
Entonnoir rempli de particules de sable
t = 5 mm
24
fluidisé en expansion avait une hauteur de 12.5 cm et la sonde à fibre-optique était insérée dans le
lit à une hauteur de 7 cm au-dessus du distributeur. Un miroir était positionné
perpendiculairement à l’extrémité de la sonde. Les bulles de gaz étaient produites dans le volume
de mesure de la sonde à fibres-optiques en injectant une mixture de 10.1 % CH4 + 89.9% N2 au
moyen d’un injecteur orienté vers le bas. La mixture de CH4/N2 était injectée avec un débit de
10 mL/s par l’ouverture manuelle d’une valve pendant approximativement 0.5 s et à un intervalle
d’environ 11 secondes.
Figure 3-5: Réacteur à lit fluidisé
CH4/N2
N2
Sonde à fibres-optiques IR
FT
-IR Miroir plan
Bulle FCC
Plaque poreuse
25
CHAPITRE 4 COMBUSTION EN LIT FLUIDISÉ DES
N-ALKANES C1-C4
Cet article a été soumis au journal Fuel.
4.1 PRÉSENTATION DE L’ARTICLE
L’objectif de cet article est d’étudier la combustion des n-alcanes C1-C4 dans un lit fluidisé de
solides inertes. Les températures nécessaires à la conversion complète des alcanes C1-C4 dans la
région du lit fluidisé (températures critiques de transition) ont été mesurées. De plus, une
cinétique de réaction globale et de premier ordre est mesurée pour chacun des alcanes.
Finalement, un pour prédire la conversion dans la région de l’injecteur et la région du lit fluidisé
sont proposés.
Les résultats mettent en évidence des similitudes entre la combustion des n-alcanes C2-C4: ils
sont caractérisés par les mêmes températures critiques de transitions T1 (température du lit
fluidisé en-dessous de laquelle la combustion se produit essentiellement dans la zone de
désengagement) et T2 (température du lit fluidisé au-delà de laquelle la combustion se produit
entièrement au niveau du distributeur). De plus, la cinétique des n-alcanes C2-C4 peut être décrite
par une seule expression de type Arrhénius. Dans le cas du méthane, les taux de combustion
observés sont significativement plus lent de sorte que les températures critiques de transition sont
plus élevées et la cinétique de combustion plus lente.
L’article propose deux modèles réacteur piston en séries afin de prédire la conversion dans le lit
fluidisé: un modèle pour la région de l’injecteur et un modèle pour le lit fluidisé. Les résultats
démontrent qu’une hausse de la température du lit augmente la conversion au niveau de
l’injecteur, tandis qu’une élévation de la vitesse d’injection décroît la conversion. La conversion à
proximité de l’injecteur est modélisée par un réacteur piston dont la longueur est corrélée à la
température par une expression de type Arrhénius, de façon similaire au temps d’induction d’une
mixture inflammable homogène. L’effet inhibiteur de la vitesse d’injection est également inclus
dans cette corrélation.
26
FLUIDIZED BED COMBUSTION OF C1-C4 N-ALKANES
Jean-Philippe Laviolette, Gregory S. Patience, Jamal Chaouki
(Soumis à Fuel, 2010)
4.2 FLUIDIZED BED COMBUSTION OF C1-C4 N-ALKANES
4.2.1 Abstract
The non-premixed combustion of C1 to C4 n-alkanes with air was investigated inside a bubbling
fluidized bed of inert sand particles at intermediate temperatures: 923 K ≤ TB ≤ 1123 K. For
ethane, propane and n-butane, combustion occurred mainly in the freeboard region at bed
temperatures below T1 = 923 K. On the other hand, complete conversion occurred within 0.2 m
of the injector at: T2 = 1073 K. For methane, the measured values of T1 and T2 were significantly
higher at 1023 K and above 1123 K, respectively. The fluidized bed combustion was accurately
modeled with first-order global kinetics and two one-phase PFR models in series: one PFR to
model the region close to the injector and another to represent the main fluidized bed body. The
measured global reaction rates for C2 to C4 n-alkanes were characterized by a uniform Arrhenius
expression, while the global reaction rate for methane was significantly slower. Reactions in the
injector region either led to significant conversion in that zone or an autoignition delay inside the
main fluidized bed body. The conversion in the injector region increased with rising fluidized bed
temperature and decreased with increasing jet velocity. To account for the promoting and
inhibiting effects, an analogy was made with the concept of induction time: the PFR length (bi) of
the injector region was correlated to the fluidized bed temperature and jet velocity using an
Arrhenius expression.
Keywords: combustion, fluidized bed, methane, propane, ethane, n-butane and non-premixed
27
4.2.2 Introduction
In the context of increasingly more severe environmental regulations due to global warming, new
fluidized bed processes are currently under development to produce energy and chemical
products from biomass and natural gas feedstocks. Combustion is inherent to many of these
processes, which include selective oxidation of alkanes, biomass gasification, combustion of non-
conventional feedstocks (biomass and co-firing) and many more. Combustion may involve many
hydrocarbons present in the feedstock or produced from thermal cracking during the reaction. A
good knowledge of the combustion process and the operating conditions necessary for complete
combustion in the fluidized bed region is essential for process design and safety during operation,
startup and shutdown.
The combustion of methane, LPG and propane in fluidized beds of inert particles has been the
subject of several studies in premixed [1-10] and non-premixed modes [11-16]. Fluidized bed
temperature has been reported as a key parameter that determines in-bed fuel conversion: the
combustion front moves towards the distributor as the fluidized bed temperature is increased. At
low fluidized bed temperature, the combustion is first initiated in the freeboard region. Increasing
the temperature moves the combustion front to the bubbles bursting at the upper surface of the
bed, then to bubbles igniting in the bed and finally to small bubbles where ignition occurs
practically at the level of the distributor and the process appears flameless [7, 9]. While some
studies have suggested that combustion occurs predominantly inside the bubble phase as opposed
to the emulsion phase [9], methane and propane combustion have been reported in fixed beds of
sand particles at temperatures above 1023 K and 973 K, respectively [1, 17].
Two critical transition temperatures have been defined for fluidized bed combustion in inert
particles. Below the lower critical temperature (T1), combustion only occurs above the bed
surface. Between T1 and the upper critical temperature (T2), combustion begins to move inside
the bubbles within the bed. Finally, above T2, combustion takes place entirely within the bed and
close to the distributor. For methane combustion in inert sand particles, the values of T1 and T2
reported in the literature range between 823-1018 K and 1133-1250 K, respectively [18]. For
28
propane, T2 was measured in a shallow bed (HB ≈ 0.1 m) of sand particles as 1123 K [2, 9].
Investigations on the fluidized bed combustion of other alkanes in inert particles are scarce.
Fluidized bed combustion can be estimated from a reaction model that combines the gas/solids
hydrodynamics and the reaction kinetics. Several gas-phase hydrodynamics models of varying
complexity have been proposed in the literature [18]. Hydrocarbon conversion in fluidized beds
has been shown to vary from plug flow to back-mixed flow depending on the fluidization
conditions and the reaction. A recent gas tracer study by Lorences et al. [19] showed that the gas-
phase hydrodynamics is very close to plug flow when Ug ≤ 6 Umf.
The mass transfer rate at the distributor is greater than inside the fluidized bed main body [18].
For fast reactions, such as combustion, significant conversion can occur in the injector tip
surroundings. Therefore, the region near the injector and the main fluidized bed body require two
different gas-phase hydrodynamic models. For a downward-facing sparger, counterflow ignition
studies should yield some insights into the parameters affecting conversion at the injector tip [20-
22]. These studies show that the ignitability and conversion decreases with decreasing
temperature and increasing strain rate (counterflow gas velocities). The latter can be explained by
the fact that the ignition of hydrocarbons involves intermediate species whose lifetimes may be
similar to the characteristic time scales of mass and heat transfer. To characterize the injector
region, one-phase and two-phase models have been proposed in the literature [18].
Several microkinetic and first-order global kinetics schemes are available in the literature for gas-
phase systems [5, 6, 23]. Gas-phase microkinetic models have also been modified to account for
the “quenching” effect of solids particles on the free radicals [17, 24]. Microkinetic reaction
networks are complex but essential to simulate the production of intermediate species. However,
in fuel lean conditions, the parent hydrocarbon conversion inside a fluidized bed has been shown
to be accurately represented by a first order reaction [17, 25, 26]. Global gas-phase kinetics have
been proposed for methane, but they are scarce in the literature for ethane and higher alkanes
(C2+) [23].
29
It may be possible to express the first-order kinetics or conversion of C2+ alkanes in terms of a
single expression or correlation. Studies have reported similarities between the combustion
chemistry of n-alkanes: their combustion rate is controlled by the same main chain branching
path, which can be summarized by the following elementary reactions at intermediate
temperatures (~ 850 K to 1200 K) [27-30]:
MHOMOH +→++ 22 (1)
222 OHRHORH +→+ (2)
MOHOHMOH ++→+22 (3)
In the above equations, RH is an alkane, R is an alkyl radical and M is a third body.
Saxena et al. [29] have demonstrated that the high temperature induction time of propane and
higher alkanes can be accurately estimated by employing rate parameters of elementary steps (2)
to (4).
22222 OOHHOHO +→+ (4)
In fact, the induction time of C2-C5 n-alkanes measured in shock tubes shows a similar
dependence on temperature, pressure and mixture composition such that it may be calculated
from a single correlation with an uniform activation energy [27-28]. The previous studies show
that at a similar stoichiometry, temperature and pressure, the induction time decreases as C2H6 <
n-C4H10 < C3H8 [27, 28, 31]. Methane, on the other hand, has been observed to yield significantly
higher induction times, which can be explained by the production of a methyl radical in equation
(2), which recombines to form ethane, resulting in chain termination. If the combustion chemistry
and induction time correlations among C2+ n-alkanes show such similarities, a parallel may also
be made between their global reaction kinetics and conversion.
In the present study, the non-premixed combustion of methane, ethane, propane and n-butane was
studied in a fluidized bed of sand particles at temperatures ranging from 923 K to 1123 K. The
flow of each hydrocarbon was adjusted to obtain a constant heat release rate. Lower and upper
critical bed temperatures (T1 and T2) were measured for each hydrocarbon. Furthermore, a simple
30
reaction model with first-order reaction kinetics was developed and compared to the experimental
results. The first-order kinetic parameters were measured for each n-alkane and compared.
4.2.3 Experimental
The experiments were performed in the fluidized bed reactor shown in Figure 4.1. The region of
the fluidized bed of solids had an inner diameter of 0.2 m and a height of approximately 1.0 m.
Inert sand particles (dp = 290 µm and Umf = 0.08 m/s) were used as the bed material. The bed was
fluidized with air, injected through the distributor, and the hydrocarbon fuels (methane, ethane,
propane and n-butane) were injected through a downward-facing sparger with the tip located at a
distance of 0.1 m above the distributor. Prior to each experiment, the fluidized bed reactor was
heated to the desired temperature with a natural gas burner connected to the windbox. Heavy
vacuum gas oil (HVGO) was also injected through the sparger to accelerate the heating process.
Once the bed temperature had reached the desired value, the HVGO injection and the natural gas
burner were stopped and the fluidized bed was heated for about 1 hour with propane to burn any
HVGO residues.
Figure 4-1: Fluidized bed reactor
3 4
Non-isokinetic sampling probes
48 cm
20 cm
97 cm
38 cm
1 1st stage cyclone
2 2nd
stage cyclone 3 Natural gas burner 4 Windbox 5 Fluidized bed
60 cm
20 cm
Air + Natural gas
C3H8
r
5
GC
O2 analyzer
CO/CO2 analyzer
z
1
2
31
The experiments were conducted at a superficial gas velocity of 0.4 m/s and at five bed
temperatures: 923 K, 973 K, 1023 K, 1073 K and 1123 K. The temperature was measured at 9
different axial positions with thermocouples. Gas was sampled at six axial and five radial
positions. The sampled gas was analyzed by a Siemens Ultramat™ CO/CO2 analyzer and a
Varian CP-4900™ micro-GC connected in series to measure the volume fraction of the following
species: H2, CH4, C2H4, C2H6, C3H8, n-C4H10, O2, N2, CO and CO2. At each axial position, a
cross-sectional average (surface area weighted) of species volume fraction was calculated. Gas
was also sampled in the disengagement region and the oxygen volume fraction was recorded by
an oxygen analyzer. The mixed-cup volume fraction at the injector tip was adjusted for each
hydrocarbon to obtain a heat release rate of 3.8 kW (calculated from the lower heating value). For
methane, ethane, propane and n-butane, the corresponding mixed-cup volume fractions were
5.1%, 2.5%, 1.9% and 1.3%, respectively.
4.2.4 Results and discussion
4.2.4.1 Critical transition fluidized bed temperatures
Lower (T1) and upper (T2) critical transition bed temperatures were measured for methane,
ethane, propane and n-butane.
4.2.4.2 Ethane, propane and n-butane
Figure 4.2a shows the axial profile of conversion for ethane, propane and n-butane inside the
fluidized bed reactor at a bed temperature of 923 K. The corresponding axial profile of CO2
volume fraction is presented in Figure 4.2b. The expanded bed height was 0.55 m and it is shown
on both figures as a dashed line. At 923 K, ethane, propane and n-butane conversion was
insignificant inside the fluidized bed: conversion fluctuated between 0-4%, 0-18% and 10-27%
for ethane, propane and n-butane, respectively. Furthermore, the CO2 volume fractions in the
fluidized bed remained low at approximately 0.7%. Combustion occurred mainly at the bed
surface and inside the freeboard as conversion reached 92%, 93% and 96% at 0.2 m above the
32
bed surface (Z = 0.75 m) for ethane, propane and n-butane, respectively. This was accompanied
by a significant increased in the CO2 volume fraction, which reached approximately 5.5% for all
alkanes. Complete conversion was obtained at Z = 0.97 m.
As the bed temperature was increased to 973 K, the combustion front moved upstream inside the
fluidized bed for ethane, propane and n-butane. Figures 4.2c and 4.2d show the axial profile of
conversion and CO2: combustion was initiated within 0.15 m (Z = 0.25 m) of the injector tip
where conversion was 24%, 57% and 61% for ethane, propane and n-butane, respectively.
Conversion increased with height in the fluidized bed and reached 60%, 79% and 84% at 0.1 m
below the bed surface (Z = 0.46 m) for ethane, propane and n-butane. The unconverted
hydrocarbons burned inside the freeboard region and complete conversion was observed at 0.2 m
above the bed surface (Z = 0.75 m). The combustion in the fluidized bed and freeboard regions
was accompanied by a corresponding increase in the CO2 volume fraction, which reached
approximately 3.8% at 0.1 m below the fluidized bed surface (Z = 0.46 m) and 6.0% in the
freeboard region at Z = 0.75 m (complete combustion).
Figures 4.2(e) and 4.2(f) show the axial profile of CO2 and conversion for the three n-alkanes as
the bed temperature was increased further to 1023 K. Near complete fuel conversion occurred in
the fluidized bed as conversion reached 95%, 95% and 98% for ethane, propane and n-butane,
respectively, only 0.36 m (Z = 0.46 m) downstream of the injector. At that position, the CO2
volume fraction reached about 6.0%, which indicated that complete combustion was nearly
achieved. Complete conversion was observed 0.2 m over the bed surface (Z = 0.75 m). Rising the
bed temperature by 50 K to 1073 K resulted in 99% conversion within 0.25 m of the sparger tip
(Z = 0.35 m) for all three alkanes.
These results demonstrate that the fluidized bed combustion of the C2-C4 alkanes is characterized
by similar critical bed temperatures. At a bed temperature of 923 K, combustion occurred
primarily in the freeboard region for C2 to C4 n-alkanes: this was defined as the lower critical
temperature (T1). Furthermore, complete combustion occurred in the fluidized bed region and
within 0.25 m of the sparger tip at 1073 K, which was defined as the upper critical temperature
33
(T2). This is indicative of similar temperature dependence for C2 to C4 n-alkane fluidized bed
combustion and similar global activation energies.
Figure 4-2: Conversion and YCO2 axial profiles at 923 K, 973 K and 1023 K
34
The measured T2 is lower than the value of 1123 K previously reported for propane [2, 9].
However, this discrepancy may be explained by the fact that conversion was not measured at a
bed temperature of 1073 K – experiments were performed at bed temperatures of 1023 K and
1123 K [2, 9].
4.2.4.3 Methane
Compared to ethane, propane and n-butane, methane fluidized bed combustion was initiated at
significantly higher bed temperatures. Figure 4.3 shows the axial profile of methane conversion at
973 K, 1023, 1073 and 1123 K. At a bed temperature of 973 K, no methane conversion was
measured in the fluidized bed (0% at Z = 0.46 m) and combustion occurred entirely in the
freeboard region: methane conversion increased from 37% at 0.06 m above the bed surface
(Z = 0.61 m) to 99% at Z = 0.97 m, respectively.
Figure 4-3: XCH4 axial profiles at various fluidized bed temperatures
As the temperature was increased to 1023 K, the methane in-bed conversion was low – the
measured conversion fluctuated between 0% and 12%. Combustion mainly occurred at the bed
35
surface – conversion increased to 64% at 0.06 m above the bed surface (Z = 0.61 m). This was
accompanied by a significant increase in the volume fraction of CO2 in the freeboard. Complete
methane conversion was observed in the freeboard region at Z = 0.97 m.
Figure 4-4: Comparison between the model and experiments at TB = 973, 1023,1073 and 1123 K
The bed temperature was risen by 50 K to 1073 K and significant in-bed conversion was
observed: conversion reached 23% at a distance of 0.15 m downstream of the sparger
(Z = 0.25 m) and increased to 40% at 0.09 m below the bed surface (Z = 0.46 m). The
unconverted methane burned in the freeboard and complete conversion was recorded
at Z = 0.97 m.
36
The bed temperature was increased further to 1123 K and methane conversion increased to 83%
at 0.1 m below the bed surface (Z = 0.46 m). The unconverted methane burned inside the
freeboard region and conversion reached 98% only 0.06 m above the bed surface (Z = 0.61 m).
Fluidized bed combustion required significantly higher bed temperatures compared to ethane,
propane and n-butane. Methane conversion occurred essentially in the freeboard region for bed
temperatures below 1023 K, which was defined as the lower critical temperature (T1). This value
of T1 is in the range of previous observations [18]. Furthermore, complete in-bed conversion of
methane required bed temperatures above 1123 K and this agrees with the reported T2 values in
the scientific literature [18]. The measured lower and upper critical bed temperatures for
methane, ethane, propane and n-butane are listed in Table 4.1.
Table 4.1: Measured T1 and T2 for C1 to C4 n-alkanes
n-Alkane T1 (K) T2 (K)
CH4 1023 > 1123
C2H6 923 1073
C3H8 923 1073
n-C4H10 923 1073
4.2.4.4 C1-C4 n-alkanes global reaction rates
A simple one-phase reaction model was developed to predict the in-bed conversion of methane,
ethane, propane and n-butane. The fluidized bed gas-phase hydrodynamics was modeled as a
plug flow reactor (PFR) and first-order kinetics was used. The fluidized bed reactor was
separated in two separate zones: (1) the injector region (high gas/gas and gas/solid mixing near
37
the sparger tip) and (2) the main fluidized bed body. The hydrocarbon conversion was calculated
from equations (4.5) to (4.7):
BedFluidized,iInjector,ii XXX += (4.5)
( )( )ij bZk
i eX+′−
−= 1100 (4.6)
RTE
RTE AA
ekev
kAk
−−′==′ 0
0 (4.7)
The parameter Zj corresponds to the position where the fuel/air mixing occurs and where the free
radical build-up is initiated. To determine Zj, the flow pattern at the injector tip was characterized
by calculating the jet penetration depth using six correlations [32-37]. Jet penetration depth
correlations for downward injectors are scarce in the literature. However, previous observations
have shown that the jet penetration depth is similar for horizontal and downwards jets [38].
Table 4.2 lists the jet penetration depths calculated from three correlations: the obtained values
were small and comprised between 0.0 m and 0.034 m. Three correlations yielded negative jet
penetration depths, which may be explained by the fact that the present operating conditions are
outside the validity range of the correlations [31-33]. The three correlations are not included in
Table 4.2. Since the calculated jet penetration depths were small, Zj = 0.0 m was taken as the
injector tip position (Z = 0.1 m).
Figure 4.4a shows the in-bed conversion of ethane, propane and n-butane as a function of Zj on a
natural logarithmic scale for two bed temperatures: 973 K and 1023 K. A linear relationship is
observed between the ordinate [ln(1-Xi)] and Zj, which indicates that the model agrees well with
the experimental results. The slope of the curve represents the first order reaction rate (k’): the
same reaction rate was measured for ethane, propane and n-butane. For bed temperatures of
973 K and 1023 K, k’ was measured as 3.7 m-1 and 8.2 m-1, respectively. Therefore, the
temperature dependence of the reaction rate was uniform for C2 to C4 n-alkanes, which implies
that the activation energy (EA) was also the same.
Similar observations were made at higher bed temperatures. Figure 4.4b shows the in-bed
conversion of methane, ethane and propane as a function of Zj on a natural logarithmic scale for
38
two bed temperatures: 1073 K and 1123 K. The first-order global reaction rate for ethane and
propane was measured as 13.5 m-1 at 1073 K. N-butane conversion reached 100% and could
therefore not be included in Figure 4.4b. For methane, however, the reaction rate was
significantly lower compared to C2-C4 n-alkanes: k’ rose from 1.2 m-1 to 4.2 m-1 as the bed
temperatures was increased from 1073 K to 1123 K.
Table 4.2: Jet penetration depth correlations
Correlation Jet penetration (cm)
Orientation Ref. CH4 C2H6 C3H8 C4H10
( ) 21
87j
j
p
j
j
j
gD
U.
D
L
=
ρ
ρ 0.06 0.06 0.06 0.06 Horizontal [32]
( )
2702
525
.
jgp
jj
j
j
gD
U.
D
L
−=
ρρ
ρ 3.4 2.8 2.6 2.4 Horizontal [33]
21
2
=
b
gnj w
QL
ρ
ρ 1 0.7 0.6 0.4 Downward [34]
The global reaction rates can be expressed in the Arrhenius form:
( ) RTCH e.k
251008124 10991
−×=′ (R = 8.314 J/mol.K) (4.8)
( ) RTHCHCHC e.kkk
12730071048362 10592
−×=′=′=′ (4.9)
The reaction rate constant and the activation energy for methane were calculated from the
reactions rates measured at bed temperatures of 1073 K and 1123 K. On the other hand, the
parameters for ethane, propane and n-butane were calculated from the reaction rates at 973 K and
1023 K. Equation (4.9) overpredicts the global reaction rate at 1073 K and yields 16.4 m-1
compared to the measured value of 13.5 m-1. This discrepancy can be explained by the fact that
the in-bed conversions were very high at this temperature (in the order of 93% to 100%). At these
39
high conversions, a variation of 1.0% in the measured value greatly affects the measured slope in
Figure 4.4b. The precision in the conversion measurements was estimated as ±5% absolute from
the variations in repeated measurements.
The fact that ethane, propane and n-butane have the same global kinetics is consistent with
previous studies, which have shown that the induction times of C2 and higher alkanes are of
similar magnitude and are governed by the same temperature dependence or activation
energy [27-28]. Methane is also known to be significantly less reactive than higher
n-alkanes [28, 30].
4.2.4.5 Contribution from the injector region
The region close to the injector is modeled as a separate PFR and the parameter bi of equation
(4.7) represents its contribution to the overall fluidized bed conversion. A positive value for bi
means that combustion occurred in the injector region. On the other hand, a negative value for bi
implies that autoignition did not occur in the injector region and occurred in the main fluidized
bed body only after an induction delay. Therefore, the parameter bi is analogous to an induction
distance as it increases with decreasing induction delay.
Table 4.3: Parameter bi at the different bed temperatures (TB)
n-Alkane
Parameter bi (m) for various TB (K)
973 1023 1073 1123
CH4 N/A N/A -0.075 0.05
C2H6 -0.085 -0.020 0.010 N/A
C3H8 0.050 0.050 0.06 N/A
n-C4H10 0.100 0.100 0.100 N/A
40
The values for bi were determined such that the model would fit with the experimental data.
Table 4.3 lists the values of bi for C1-C4 n-alkanes for all tested fluidized bed temperatures. For
methane and ethane, bi decreased with temperature, which is consistent with previous induction
time studies where induction time decreased with increasing temperature. For methane, bi
increased from -0.075 m at 1073 K to 0.05 m at 1123 K. For ethane, bi increased from -0.085 m
at 973 K to 0.01 m at 1073 K. For propane and n-butane, bi remained approximately constant at
0.05 m and 0.1 m, respectively.
Figures 4.4c and 4.4d compare the measured and predicted conversion of C1 to C4 n-alkanes for
fluidized bed temperatures of 973 K to 1123 K: the agreement between the model and the
experimental results is good.
Figure 4-5: Conversion in the injector region
To better understand the observed trends in the values of bi in Table 4.3, Figure 4.5a shows the
conversion (Xi) measured 0.15 m (Z = 0.25 m) above the injector tip as a function of fluidized
bed temperature (TB) and fuel injection velocity (Uj) for C2H6, C3H8 and n-C4H10. Note that the
injection velocity at the sparger tip varied from 23-24 m/s, 18-19 m/s and 12-13 m/s for C2H6,
C3H8 and n-C4H10, respectively. For a specific bed temperature, the lowest conversion was
obtained with ethane, followed by propane and n-butane. However, these hydrocarbons share
similar kinetics as demonstrated earlier and autoignition temperature measured in non-premixed
41
counterflow experiments have been shown to increase as C2H6 < n-C4H10 < C3H8 [20-22,
27, 28, 30]. Therefore, this strongly suggests that the decrease in conversion for ethane is due to
the increased injection velocity and strain rate, which quench intermediate species and free
radicals near the injector. It is also observed from Figure 4.5a that the reduction in conversion
due to an increase in fuel injection velocity decreases with rising bed temperature. At higher
temperatures, the reaction rates may become so fast that they remain relatively unaffected by the
enhanced mixing near the sparger tip.
Since the parameter bi is analogous to an induction distance, it was attempted to produce a
correlation of bi based on the Arrhenius equation. Induction time has been shown in several
studies to be well correlated to temperature through an Arrhenius-type expression. Similarly to
fluidized bed temperature, it was assumed that the fuel injection velocity at the sparger had an
exponential effect on bi:
bi = (Average distance travelled by the gas in injector region) – (induction distance) (4.10)
( ) ( )ε
τε −
−−
∝11
ggji
UUtb (4.11)
B
injRT
U
i e..b582
0438020830 −= (4.12)
Figure 4.5b shows the parameter bi as a function of
RT
Uexp inj580
: the correlations is very
good at reproducing the experimental trend.
As shown in equation (4.11), the parameter bi is proportional to the average gas residence time in
the injector region (tj) and the mixture induction time (τ) through the interstitial gas velocity. It
seems reasonable to believe that the autoignition in the non-premixed hydrocarbon/air mixture is
governed by the induction time at the stoichiometry concentration. In fact, the hydrocarbon/air
interface at the sparger should be characterized by a wide range of equivalence ratios and
42
autoignition should be determined by the conditions that are most reactive. The average fluidized
bed solids fraction (1-ε) can be estimated from the pressure drop across the bed (5.5 kPa), the
solids density (ρp = 2650 kg/m3) and the bed height (HB = 0.55 m): (1-ε) = 0.39. Using this value
of solids fraction results in an average gas residence at the injector (tj) of 0.200 s and an induction
time that decreased from 0.160 s to 0.140 s for propane (Uj = 18 m/s) as the bed temperature was
increased from 973 K to 1073 K. In comparison, the low-temperature induction time correlation
of Penyazkov et al. [39] yields lower values of 0.033 s at 973 K and 0.012 s at 1073 K for a
stoichiometric propane/air mixture. The lower values of Penyazkov et al. [39] may be explained
by the fact that the measurements were performed on a homogeneous mixture in a different
apparatus (shock tube): a non-premixed mixture and strain rates increase induction time in the
present study.
4.2.5 Conclusions
The non-premixed combustion of methane, ethane, propane and n-butane with air was
investigated in a fluidized bed of inert sand particles. The bed temperature was varied between
923 K and 1123 K while the superficial gas velocity was kept constant at 0.4 m/s. Critical
transition bed temperatures (T1 and T2) were measured for each fuel. For ethane, propane and
n-butane, combustion occurred mainly in the freeboard region at bed temperatures below
T1 = 923 K (650oC). On the other hand, complete conversion occurred within 0.2 m of the
injector at: T2 = 1073 K. For methane, the measured values of T1 and T2 were significantly higher
at 1023 K and above 1123 K, respectively.
The fluidized bed combustion was accurately modeled with first-order global kinetics and two
one-phase PFR models in series: one PFR to model the region close to the injector and another to
represent the main fluidized bed body. The global reaction rates of C2 to C4 n-alkanes were
evaluated and found to be characterized by a uniform Arrhenius expression. On the other hand,
the methane global reaction rate was significantly slower with a higher activation energy.
43
The reaction model for the injector region either increased conversion near the sparger tip or
delayed autoignition inside the main fluidized bed body. The conversion in the injector region
increased with rising fluidized bed temperature and decreased with increasing jet velocity (strain
rate). To account for the promoting and inhibiting effects in the sparger region, an analogy was
made with the concept of induction time: the PFR length (bi) was varied as a function of the
fluidized bed temperature and jet velocity. The PFR length was correlated to these two
parameters using an Arrhenius expression, similarly to induction time correlations.
4.2.6 Nomenclature
A Cross-sectional area (m2)
bi Parameter in equation (7) for ith species (m)
Dj Injector nozzle diameter (m)
EA Activation energy (J)
g Gravitational constant (9.81 m/s2)
HB Fluidized bed height (m)
k0 First-order reaction rate constant (m-1)
Lj Jet penetration length (m)
Qn Jet volumetric flow rate at 0oC & 101.3 kPa (litres)
R Universal gas contant (8.314 J/Kmol)
r Radial position (m)
tj Average gas residence time in injector region (s)
T Temperature (K)
TB Fluidized bed temperature (K)
T1 Lower critical temperature (K)
T2 Upper critical temperature (K)
44
Uj Gas injection velocity at the sparger tip (m/s)
Umf Minimum fluidization velocity (m/s)
Ug Superficial gas velocity (m/s)
v Volumetric flow rate (m3/s)
w Nozzle thickness (cm)
Xi Conversion of ith species (%)
Yi Volume fraction of ith species (vol%)
Z Axial position above the distributor (m)
Zj Axial position above the injector tip (m)
Greek letters
ε Average void fraction in fluidized bed
ρb Fluidized bed density (kg/m3)
ρg Fluidizing gas density (kg/m3)
ρj Gas density at injector nozzle (kg/m3)
ρp Particle density (kg/m3)
τ Induction time (s)
4.2.7 References
[1] Hesketh RP, Davidson JF. Combustion of methane and propane in an incipiently fluidized
bed. Combust Flame 1991;85:449-467.
[2] Van der Vaart DR. Freeboard ignition of premixed hydrocarbon gas in a fluidized bed.
Combust Flame 1988;71:35-39.
45
[3] Van der Vaart DR. The chemistry of premixed hydrocarbon/air combustion in a fluidized bed.
Fuel 1988;67:1003-1007.
[4] Friedman J, Li H. Natural gas combustion in a fluidized bed heat-treating furnace.
Combustion Sci Technol 2005; 177; 2211-2241.
[5] Dounit S, Hemati M, Andreux R. Modelling and experimental validation of a fluidized-bed
reactor freeboard region: application to natural gas combustion. Chem Eng J 2008; 140; 457-465.
[6] Pre P, Hemati M, Marchand B. Study on natural gas combustion in fluidized beds: modelling
and experimental validation. Chem Eng Sci 1998; 53 (16); 2871-2883.
[7] Zukowski W. A simple model for explosive combustion of premixed natural gas with air in a
bubbling fluidized of inert sand. Combust Flame 2003; 134; 399-409.
[8] Baron J, Bulewicz EM, Zukowski W, Kandefer S, Pilawska M. Combustion of hydrocarbon
fuels in a bubbling fluidized bed. Combust Flame 2002; 128; 410-421.
[9] Dennis JS, Hayhurst AN, Mackley IG. The ignition and combustion of propane/air mixtures
in a fluidised bed. 19th Symp. (Intl.) on Combust. 1982; 1205-1212.
[10] Srinivasan RA, Sriramulu S, Kulasekaran S, Agarwal PK. Mathematical modeling of
fluidized bed combustion-2: combustion of gases. Fuel 1998; 77; 1033-1049.
[11] Mabrouk R, Chaouki J, Laviolette J-P, Patience GS. On natural gas combustion in a
fluidized bed. Proc. 9th Conference on Circulating Fluidized Beds, Hamburg, Germany, 2008.
[12] Ross DP, Yan HM, Zhang DK. An experimental study of propane combustion in a fluidized-
bed gasifier. Combust Flame 2001; 124; 156-164.
[13] Sotudeh-Gharebaagh R. Combustion of natural gas in a turbulent fluidized bed reactor. PhD
Thesis, Ecole Polytechnique, Montreal, Canada, 1998.
[14] Singh B, Rigby GR, Callcott TG. Combustion of gaseous fuels in a fluidised bed. Fluidised
Combustion Conference, Paper C5, London, 1975.
[15] Stubington JF, Davidson JF. Gas-phase combustion in fluidized beds. AIChE Journal 1981;
27 (1); 59-65.
46
[16] Sotudeh-Gharebaagh R, Chaouki J. Investigation of highly exothermic reactions in a
turbulent fluidized bed reactor. Energy & Fuels 2007; 21; 2230-2237.
[17] Sotudeh-Gharebagh R, Chaouki J. The heterogeneous and homogeneous combustion of
methane over inert particles. Can J Chem Eng 2003; 81; 1182-1191.
[18] Laviolette J-P, Sotudeh-Gharebagh R, Mabrouk R, Patience GS, Chaouki J. In A. Agarwal
(ed.)Handbook of Combustion (vol.5), Wiley 2010.
[19] Lorences M, Laviolette J-P. Fluid bad gas RTD: effect of fines and internals. Powder
Technol 2006; 168 (1); 1-9.
[20] Humer S, Seiser R, Seshadri K. Non-premixed and premixed extinction and autoignition of
C2H4, C2H6, C3H6 and C3H8. P Combust Inst. 2002; 29; 1597-1604.
[21] Fotache CG, Wang H, Law CK. Ignition of ethane, propane, and butane in counterflow jets
of cold fuel versus hot air under variable pressures. Combust Flame. 1999; 117; 777-794.
[22] Jomaas G, Zheng XL, Zhu DL, Law CK. Experimental determination of counterflow
ignition temperatures and laminar flame speeds of C2-C3 hydrocarbons at atmospheric and
elevated pressures. P Combust Inst. 2005; 30; 193-200.
[23] Ross DP, Yan H-M, Zhang D-K. Modelling of a laboratory-scale bubbling fluidised-bed
gasifier with feeds of both char and propane. Fuel 2004; 83; 1979-1990.
[24] Jeng R-S, Altwicker ER, Morgan III MM, Wilk RD. Propane combustion in a spouted bed
combustor II: a modeling study in the bed region. Combustion Science and Technol 2001; 170;
119-149.
[25] Foka M, Chaouki J, Guy C, Klvana D. Natural gas combustion in a catalytic turbulent
fluidized bed. Chem. Eng. Sci. 1994; 49 (24A); 4269-4276.
[26] Chaouki J, Klvana D, Guy C. Selective and complete catalytic oxidation of natural gas in
turbulent fluidized beds. Korean J. Chem. Eng. 1999; 16 (4); 494-500.
[27] Horning DC, Davidson DF, Hanson RK. Study of the high-temperature autoignition of
n-alkane/O2/Arixtures. Journal of Propulsion and Power 2002; 18 (2); 363.
[28] Burcat A, Scheller K, Lifshitz A. Shock-tube investigation of comparative ignition delay
times for C1-C5 alkanes. Combust Flame 1971; 16; 29-33.
47
[29] Saxena P, Peters N, Williams FA. An analytical approximation for high-temperature
autoignition times of higher alkanes. Combust Flame 2007; 149; 79-90.
[30] Westbrook, CK. Chemical kinetics of hydrocarbon ignition in practical combustion systems.
Proc Combust Inst 2000; 28; 1563-1577.
[31] Lamoureux N, Paillard C-E, Vaslier V. Low hydrocarbon mixtures ignition delay times
investigation behind reflected shock waves. Shock Waves 2002; 11; 309-322.
[32] Zenz FA, Bubble formation and grid design, IChemE Symposium Series 1968; 30; 136–139.
[33] Benjelloun F, Liegeois R, Vanderschuren J. Penetration length of horizontal gas jets into
atmospheric fluidized beds, in: J.F. Large, C. Laguerie (Eds.), Fluidization VIII: Proceedings of
the Eighth Engineering Foundation Conference on Fluidization, Amer Inst of Chemical
Engineers, 1995, 239–246.
[34]Shen Z, Briens CL, Kwauk M, Bergougnou MA. Study of a downward gas jet in a two-
dimensional fluidized bed. Can J Chem Eng 1990; 68; 534-540.
[35] Shakhova NA, Discharge of turbulent jets into a fluidized bed, Journal of Engineering
Physics and Thermophysics 1968; 14 (1); 32–36.
[36] Merry JMD, Penetration of a horizontal gas jet into a fluidised bed, Trans Inst Chem Eng
1971; 49 (4); 189–195.
[37] Hong R, Li H, Li H, Wang Y. Studies on the inclined jet penetration length in a gas–solid
fluidized bed, Powder Technol 1997; 92; 205–212.
[38] Okhotskii VB. Depth of penetration of a gas jet into a granular medium. Theor Found Chem
Eng 2001; 35(2): 199-200.
[39] Penyazkov OG, Ragotner KA, Dean AJ, Varatharajan B. Autoignition of propane-air
mixtures behind reflected shock waves. Proc Combust Inst 2005; 30: 1941-1947.
48
CHAPITRE 5 COMBUSTION DU PROPANE DANS LA ZONE DE
DÉSENGAGEMENT D’UN LIT FLUIDISÉ
Cet article a été soumis au journal Fuel.
5.1 PRÉSENTATION DE L’ARTICLE
L’objectif de cet article est d’étudier la combustion du propane dans la zone de désengagement
d’un réacteur à lit fluidisé. Des expériences sont menées avec une faible vitesse superficielle de
gaz afin de minimiser l’emportement de solides en aval du lit fluidisé. Durant les expériences, le
flux de solides, la composition chimique de la phase gazeuse et la température sont mesurés
simultanément avec une sonde à prélèvement non-isocinétique et des thermocouples.
Les profiles axiaux de fractions volumétriques de propane sont comparés aux prédictions de six
modèles de microcinétique pour la combustion homogène en phase gazeuse. Les résultats
démontrent que les modèles surestiment le temps d’induction dans la zone de désengagement et
sous-estiment le taux de combustion. Or, ceci suggère que des radicaux libres dans la région du
lit fluidisé peuvent être entraînés en aval où ils accélèrent les réactions de combustion dans la
zone de désengagement.
49
GAS-PHASE PROPANE COMBUSTION IN THE FREEBOARD OF A FLUIDIZED BED
Jean-Philippe Laviolette, Gregory S. Patience, Concetta La Marca, Jamal Chaouki
(Submitted to Fuel, 2010)
5.2 GAS-PHASE COMBUSTION IN THE FREEBOARD OF A
FLUIDIZED BED
Propane combustion experiments were conducted in the freeboard of a fluidized bed of sand
particles at temperatures between 818 K and 923 K and at superficial gas velocity twice the
minimum fluidization velocity. The freeboard region was characterized by simultaneous
measurements of solids flux, chemical composition, temperature and pressure. Autoignition was
only recorded within 0.06 m of the bed surface at temperatures greater than 833 K. Propane
conversion predicted by six different microkinetic mechanistic models were compared to the
experimental measurements: all six models underestimated the reaction rate above the bed
surface. However, accounting for the production of H2O2 during in-bed combustion significantly
increased the calculated reaction rates and resulted in a better agreement between predicted and
measured propane conversion.
5.2.1 Introduction
Gas-phase homogenous combustion is a major constraint in attaining high yields, particularly
selectivity, in processes involving heterogeneous catalysis and specially as applied to partial
oxidation reactions. Together with product yield losses, combustion represents a significant
safety hazard in the case of reactions involving molecular oxygen and hydrocarbons. Chemical
processes requiring molecular oxygen are widespread in the industry. Also, new processes are
currently under development for chemicals production and power generation in the context of
increasing oil prices and global warming: selective oxidation of alkanes, biomass gasification,
and combustion of non-conventional feedstocks, for example. Fluidized bed reactors are currently
being developed for several of these processes due to their high heat transfer characteristics.
50
In order to increase productivity, fluidized bed processes may operate within the explosion
envelope while feeding the oxidant and hydrocarbon separately into the bed. This can be
achieved due to the ability of the solids phase to suppress the homogeneous reactions
(combustion) and at the same time promote selective heterogeneous reactions. However, due to
operational constraints, the processes may be operated under conditions of partial conversion. In
the case of selective oxidation reactions, for example, the selectivity has been shown to decrease
at high conversion due to the increase in the rate of product decomposition [1-4]. Therefore,
downstream of the bed – in the freeboard, cyclones and associated piping – the effluent gas phase
composition is potentially explosive, characterized by high hydrocarbon and oxygen
concentrations, elevated temperatures and an insufficient solids volume fraction to quench non-
selective homogeneous reactions [5-6]. Minimizing the risk of gas-phase combustion-
deflagration in these regions remains an important design issue.
Several explosion criteria, such as explosion limits, autoignition temperatures and induction time
correlations [7-17] have been published in the scientific literature for gas-phase homogeneous
systems. Hesketh et al. [18] showed that induction time correlations can be adequate to predict
ignition delays above a fixed bed with TB ≤ 973 K and with negligible combustion in the bed.
However, extrapolating these correlations to fluidized bed reactors introduces uncertainties since
the entrained solid particles inhibit gas-phase reactions downstream of the fluidized bed.
Furthermore, the axial and radial distribution of solids above the bed surface is non-
homogeneous: solids entrainment decreases exponentially from the bed surface until it reaches a
steady value above the transport disengagement height [19-20]. Together with heterogeneous
solids distribution, axial and radial temperature gradients may exist in the freeboard region as
well as species concentration gradients. Finally, partial reactant conversion in the fluidized bed
region produces reaction intermediates and free radicals that may influence the reaction kinetics
in the freeboard region.
To accurately predict induction time and the rate of combustion of a hydrocarbon/oxygen mixture
in the freeboard for a wide range of operating conditions, a combustion model that combines the
gas/solids hydrodynamics, reaction microkinetics, temperature gradient and correct initial
51
boundary condition must be developed. Steady state freeboard reaction models have been
developed by De Lasa et al. [21], Chen et al. [22], Walsh et al. [23],
Sotudeh-Gharebagh et al. [24], Dounit et al. [25], and Hartman et al. [26] for various reactions.
However, these models have relied on global kinetic mechanisms, and were unable to model the
free radical chemistry and induction time. Furthermore, they used correlations to estimate solids
volume fraction, which may introduce significant errors.
Elementary step level free-radical kinetics models are necessary to characterize induction time
and to account for the effect of intermediate products on reaction times. Several homogeneous
gas-phase microkinetic models for propane oxidation are available in the literature for the low
(T < 600 K) [27], intermediate (650-1000K) [28-29] and high temperature (T > 1000 K) regions
[30-33]. Some of these models have been previously modified to account for the quenching effect
of solid particles on free radicals [34-35]. A gas-phase microkinetic model was previously used
by Hutchenson et al. [4] to predict the oxygen conversion in the freeboard of a lab-scale fluidized
bed (0.128 m I.D.) of vanadyl pyrophosphate (VPP) catalyst during n-butane selective oxidation.
The calculated conversions were compared to experimental measurements: the predicted
conversions were significantly lower than the measured values. Two possible explanations were
given for this discrepancy: (1) some important heterogeneous/homogeneous reactions may not
have been included in the model and that (2) the production of free radicals in the bed was not
taken into account. However, this study relied on a limited characterization of the freeboard
region: temperature and species volume fractions were measured at one position such that the
freeboard was assumed to be perfectly mixed.
The reaction model should be better validated with experimental data of hydrocarbon combustion
in the freeboard of a fluidized bed if the parameters that affect the reaction are carefully
characterized, namely the mixture composition, solids volume fraction and temperature.
Furthermore, since these parameters are coupled, they should be measured simultaneously in the
freeboard, even though this task is very challenging.
52
In the present study, non-premixed propane fluidized bed combustion experiments were
conducted in a pilot-scale reactor (I.D. = 0.2 m) at low bed temperatures (818 K ≤ TB ≤ 923 K),
which resulted in low in-bed conversion and autoignition in the freeboard region. Sand particles
(dp = 290 µm) and a low superficial gas velocity of 0.17 m/s (Ug = 2.1×Umf) were used for all
experiments to limit solid entrainment into the freeboard region. The steady-state combustion
process in the freeboard region was characterized by simultaneous measurements of solids
entrainment flux, chemical composition, pressure and temperature. A gas-phase reaction model
with elementary step level free-radical kinetics was developed and compared to the experimental
data. Six well-established microkinetic models from the literature were used [28-33].
5.2.2 Experimental
The experiments were conducted in a fluidized bed reactor with an inner diameter of 0.2 m in the
bed and freeboard regions. Sand particles (Geldart group B, ρs = 2650 kg/m3, dp = 290 µm and
Umf = 0.08 m/s) were used as the bed material and air was the fluidizing media. Propane was
injected separately through a downward-facing sparger whose tip was located at 0.1 m above the
distributor. For all experiments, a low superficial gas velocity of 0.17 m/s (Ug = 2.1×Umf) and
low bed temperatures (818 K ≤ TB ≤ 923 K) were used. The fluidized bed surface and freeboard
region were characterized by simultaneous measurements of solids entrainment flux, chemical
composition and temperature. Solids and gas were sampled simultaneously from the reactor using
a non-isokinetic probe. The solids flux was measured from the mass of the solids samples and the
gas was analyzed with a gas chromatograph and a CO/CO2 analyzer for the following chemical
species: C3H8, C3H6, C2H6, C2H4, CH4, CO, CO2, N2, O2 and H2. Temperature was measured with
10 thermocouples positioned along the reactor’s length.
5.2.3 Kinetic modeling in the freeboard
A gas-phase reaction model with elementary step level free-radical kinetics was developed and
compared to the experimental data. Propane is known to undergo a transition from a low
temperature reaction regime (T < 600 K) to an intermediate reaction regime (T > 650 K)
53
separated by a negative temperature coefficient regime (600 - 650 K). Cool flames have been
observed to occur in the low and transition temperature regimes, the latter being attributed to a
change in the main chain branching agent from hydroperoxides to hydrogen peroxide [36-37].
The present laboratory experiments were performed with freeboard temperatures located in the
intermediate temperature region, where the main chain branching reaction is thought to be the
dissociation of hydrogen peroxide into two hydroxyl radicals [36-37]. To model the kinetics, six
well-established microkinetic reaction schemes from the literature were used and compared: the
mechanisms of Dagaut et al. [29], Marinov et al. [28], San Diego [32], GRI release 3 [33],
Sung et al. [31] and Qin et al. [30].
The model of Dagaut et al. [29] (50 species and 274 elementary reactions) has been validated
with propane oxidation experiments in a jet-stirred reactor for a wide range of equivalence ratios
(0.15 - 4) and temperatures (900 - 1200 K)., The model of Marinov et al. [28] (126 species and
638 elementary reactions) has demonstrated good agreement with NO and hydrocarbon oxidation
experiments in an atmospheric flow reactor and a temperature range of 600 to 1100 K. This
kinetic scheme includes the flame chemistry for hydrogen, methane, ethylene, ethane, propane
and ethanol. The San Diego mechanism [32] (46 species and 235 reactions), the third release of
the GRI mechanism [33] (53 species and 325 reactions), the model of Sung et al. [31] (92 species
and 621 reactions) and the model of Qin et al. [30] (70 species and 463 reactions) were also used.
These four models have been compared to experimental data (induction times, flame speeds and
species axial gradients) obtained at higher temperatures than those used in the present
investigation. However, these models contain the reaction steps for the intermediate temperature
reaction regime.
The gas-phase hydrodynamics in the freeboard were modeled as an ideal plug flow reactor (PFR).
The steady-state temperature profiles that were measured experimentally were used in the model
and the simulations were run with the ChemKin 4.1 software. The kinetic schemes excluded
heterogeneous reactions steps that could have occurred at the surface of the solid particles.
54
Two sets of simulations were performed. In the first set, the freeboard region was modelled as a
plug flow reactor and the mixture composition measured at the bed surface served as the inlet
boundary condition. In the second set of simulations, a second PFR was added upstream of the
freeboard PFR to model the fluidized bed and the production of free radicals and branching
agents in that region. The mixed-cup mixture composition at the sparger tip was considered as the
inlet boundary condition of the fluid bed PFR and the gas residence time was adjusted
individually for each model to obtain the correct in-bed conversion. The mixture at the exit of the
fluid bed PFR was fed to the inlet of the freeboard PFR.
5.2.4 Results and discussion
The first part of this investigation consisted of conducting propane fluidized bed combustion
experiments at low bed temperatures and low superficial gas velocity to observe freeboard
combustion with low solids entrainment.
5.2.4.1 Freeboard combustion experimental results
During the experiments, upward solids flux and propane volume fraction were measured
simultaneously at several axial and radial positions. The local solids entrainment flux varied with
radial position – the upward flux reached a maximum at the reactor centerline and decreased
close to the wall. Figure 5.1 shows the axial profile of the solids entrainment flux (GSU) cross-
sectional average for two bed temperatures (TB): 853 K and 923 K. The entrainment flux
decreased exponentially with height (Z) from an asymptote located at a height of 0.29 m above
the distributor (Z = 0). The asymptote position was defined as the bed surface and the beginning
of the freeboard region [20] and this was also confirmed by pressure measurements: pressure
decreased in the fluidized bed and reached a constant value a few centimetres upstream of
Z = 0.29 m. Solids entrainment flux decreased exponentially to zero at Z = 0.4 m and the
transport disengagement height (TDH) was therefore 0.11 m for both bed temperatures. The bed
temperature did not have a significant effect on the bed height and TDH in the tested range.
55
Figure 5-1: YC3H8 and GSU axial profiles in the freeboard region
Figure 5.1 also shows the axial profile of propane and carbon dioxide volume fractions (Yi) for
the two bed temperatures. A radial profile of propane volume fraction was observed at the bed
surface (Z = 0.29 m) – propane volume fraction reached a maximum of 3.1 % at the centerline
and decreased to 1.35 % at a distance of 0.02 m from the wall at a bed temperature of 853 K. A
cross-sectional average of YC3H8 is therefore presented for Z = 0.29 m in Figure 1. However,
Z (m)
0.0 0.1 0.2 0.3 0.4 0.5 0.6
Yi (
vo
l%)
0
1
2
3
4
5
6
7
GS
U (
kg/m
2s)
0
5
10
15
20
YC3H8
YCO2
GSU
TB = 853 K
Ug = 0.17 m/s
HB
(a)
Z (m)
0.0 0.1 0.2 0.3 0.4 0.5 0.6
Yi (
vo
l%)
0
1
2
3
4
5
6
7
GS
U (
kg/m
2s)
0
5
10
15
20
YC3H8
YCO2
GSU
(b)
HB
YC3H8,MC
= 1.95 %
TB = 923 K
Ug = 0.17 m/s
YC3H8,MC
= 1.95 %
56
above the bed surface (Z > 0.29 m), YC3H8 was uniform across the reactor radius, which indicated
that the freeboard was well-mixed radially at all axial positions.
As shown in Figure 5.1, the propane conversion inside the bed of solids increased with rising bed
temperature. For an inlet mixed-cup propane volume fraction (YC3H8,MC) of 1.95 vol% at the
sparger tip, the propane volume fraction decreased to 1.69 vol% (XC3H8 = 13.1 %) and 0.64 vol%
(XC3H8 = 67.1 %) at the bed surface for bed temperatures of 853 K and 932 K, respectively. The
in-bed combustion of propane was accompanied by an increase in the CO2 volume fraction from
0 % at the sparger to 0.95% and 2.00% at the bed surface for bed temperatures of 853K and
923K, respectively. The low in-bed conversion measurements are consistent with previous
studies that indicate that propane combustion is triggered near the bed surface and in the bubble
phase at these low bed temperatures [5-6].
Figure 5-2: Axial temperature profile inside the fluidized bed and freeboard regions
Propane combustion continued downstream of the bed surface, in the freeboard, as shown in
Figure 5.1. Propane conversion increased further at both temperatures just 0.06 m above the bed
surface. Propane volume fraction decreased from 1.69 vol% (XC3H8 = 13.1%) to 0.77 vol%
Z (m)
0.0 0.2 0.4 0.6 0.8 1.0
T (
K)
650
700
750
800
850
900
950
923
853
833
818
∆∆∆∆TMAX
HB
yCO2
(%)
4.3
3.00.50.0
31
69360
TB (K)
57
(XC3H8 = 60.3%) and from 0.64 vol% (XC3H8 = 67.1%) to 0.24 vol% (XC3H8 = 87.7%) for a bed
temperature of 853 K and 923 K, respectively. For both bed temperatures, propane combustion
continued downstream until nearly all the propane was consumed. At TB = 853 K, the propane
volume fraction declined to 0.03% (XC3H8 = 96.6%) at a distance of 0.31 m above the bed surface
(Z = 0.6 m). However, at the higher bed temperature, the propane combustion process reached
completion within 0.16 m of the bed surface where YC3H8 was measured as 0.02 %
(XC3H8 = 99.1 %). Propane combustion was accompanied by an increase in YCO2, which reached
about 6% for both bed temperatures when all the propane was consumed. It is interesting to note
that most of the propane combustion occurred within 0.11 m of the bed surface and in the
presence of solid particles, which have been shown to inhibit gas-phase reactions.
Propane combustion produced a significant temperature increase in the freeboard. Figure 5.2
shows the steady state axial temperature profile for five bed temperatures (TB): 923, 853, 833 and
818 K. Propane combustion in the reactor was observed at three of these bed temperatures of
923 K, 853 K and 833 K. At TB = 818 K, the fluid bed and freeboard temperatures were
insufficient to trigger autoignition. When propane combustion occurred in the reactor, the bed
temperature remained approximately constant due to the low in-bed propane conversion and the
high concentration of solid particles, which are characterized by a high thermal capacity. On the
other hand, the temperature increased significantly in the freeboard and above TB at several axial
positions. Figure 5.2 also includes the maximum temperature increase recorded (∆TMAX) and the
CO2 volume fraction (YCO2) measured 0.06 m above the bed surface. The maximum temperature
increase recorded in the freeboard during propane combustion was significantly lower than the
temperature calculated for a constant pressure adiabatic case. For example, at TB = 853 K, a
maximum temperature increase of 69 K was measured compared to a temperature increase of
968 K for constant pressure adiabatic combustion. This discrepancy can be explained by heat
losses at the reactor walls, by the presence of solids in the freeboard and by the fact that the
combustion process occurred over a distance of at least 0.1 m. The ∆TMAX measured in the
freeboard increased as the bed temperature decreased from 923 K to 853 K due to a decline in in-
bed propane conversion and a higher propane breakthrough in the freeboard. At bed temperatures
of 923 K and 853 K, propane combustion reached completion close to the bed as shown in
Figure 5.1 and as indicated by the high YCO2 measured at ZF = 0.06 m shown in Figure 5.2.
58
However, when the bed temperature was decreased to 833 K, combustion in the fluid bed and
freeboard was observed occasionally – the fluid bed and freeboard temperature approached the
threshold value at which induction time was too high and autoignition did not occur. Figure 5.2
shows the temperature profile in the reactor for TB = 833 K when combustion was observed. At
TB = 833 K, the measured ∆TMAX decreased due to an incomplete propane combustion in the
freeboard as indicated by a YCO2 of only 0.5 % at a distance of 0.06 m above the bed surface.
At TB = 818 K, combustion in the fluidized bed and freeboard was never observed: no
temperature increase and no CO2 were detected in the freeboard. The temperature in the
freeboard reached a maximum at the bed and decreased with height due to heat losses at the
reactor walls. It is interesting to note that for all cases where combustion was observed in the
freeboard (TB = 923 K, 853 K and 833 K), the onset of combustion always occurred close to the
bed and within 0.06 m of its surface. It is possible that combustion was triggered immediately at
the surface. This study was limited by the distance the between the bed surface and the first
sampling probe (0.06 m).
One series of measurements was performed at TB = 853 K during propane combustion to check
for radial temperature gradients. The thermocouples were moved along the reactor radius from
the centreline to the reactor wall and the freeboard temperature was observed to be uniform
across the freeboard radius, which suggested that the gas was well mixed radially.
5.2.4.2 Freeboard combustion modelling
A gas-phase reaction model was developed to predict induction time and the reaction rate of
propane inside the freeboard region.
5.2.4.2.1 Considering only the freeboard
The first modeling attempt consisted of using experimentally measured axial temperature profiles
(shown in Figure 5.2) in a plug flow reactor model for the freeboard portion of the reactor.
59
Preliminary gas tracer experiments were performed in the freeboard and at ambient temperature
for ZF ≥ 0.16 m: the gas-phase hydrodynamics was found to be plug flow in that region. The
stable species composition measured by the GC at the bed surface was used as the inlet boundary
condition for the freeboard and six kinetic models were used and compared.
Figure 5.3(a) shows the experimental and predicted axial profiles of propane volume fraction
(YC3H8) in the freeboard for a bed temperature of 853 K obtained with the 6 microkinetic reaction
schemes. In Figure 5.3(a), ZF = 0 m corresponds to the bed surface (Z = 0.29 m). All six reaction
mechanisms predict lower propane reaction rates compared to the experimental observations. The
GRI model [33] predicts no propane combustion in the freeboard at all. The San Diego
model [32] (SD), the model of Dagaut et al. [29] (D), the model of Sung et al. [31] (S) and the
model of Qin et al. [30] (Q) predict no propane consumption at a distance of 0.06 m above the
bed surface while the experimental data showed a 54% decrease in propane volume fraction
(from YC3H8 = 1.69% at the bed surface to YC3H8 = 0.77% at ZF = 0.06 m. At ZF = 0.31 m, these
four microkinetic models yield propane volume fractions that ranged from 1.20% (SD) to 0.69%
(Q) compared to the experimental measurement of 0.07%. The microkinetic scheme of
Marinov et al. [28] (M) fit the experimental data the best, predicting a faster propane reaction rate
inside the freeboard and a propane molar fraction of 0.20% at ZF = 0.31 m.
A similar discrepancy between the experimental and predicted propane reaction rates was
observed at a bed temperature of 923 K. Figure 5.4(a) shows the measured and predicted axial
profiles of propane volume fraction in the freeboard obtained with the six microkinetics reaction
schemes. The GRI model predict an onset of combustion at a distance of about 0.15 m above the
bed surface and a rapid consumption of propane within 0.10 m, however this trend does not fit
the experimental data very well. The other five reaction mechanisms predict an onset or propane
combustion within 0.06 m of the bed surface, however the propane reaction rate is
underestimated in all cases. At ZF = 0.16 m, the San Diego model [32] (SD) and the models of
Dagaut et al. [29] (D), Sung et al. [31] (S), Qin et al. [30] (Q) and Marinov et al. [28] (M) predict
propane volume fractions of 0.41%, 0.44%, 0.28%, 0.23% and 0.1%, respectively. These values
60
are higher than the measured value of 0.02%. Again, the model of Marinov et al. [28] (M) fitted
the experimental data the best.
Figure 5-3: Experimental and predicted YC3H8 axial profiles for TB = 853 K and Ug = 0.17 m/s:
(a) 1 PFR (freeboard) & (b) 2 PFRs in series (fluidized bed + freeboard)
(O: Experimental data, SD: San Diego [32], D: Dagaut et al. [29], S: Sung et al. [31],
Q: Qin et al. [30], M: Marinov et al. [28])
ZF (m)
0.0 0.1 0.2 0.3 0.4
YC
3H
8 (
vo
l%)
0.0
0.4
0.8
1.2
1.6
2.0
ZF (m)
0.0 0.1 0.2 0.3 0.4
YC
3H
8 (
vo
l%)
0.0
0.4
0.8
1.2
1.6
2.0
(a)M
D
GRI
Q
S
SD
TB = 853 K
M
D
QS
SD
TB = 853 K
(b)
61
Figure 5-4: Experimental and predicted YC3H8 axial profiles for TB = 923 K and Ug = 0.17 m/s:
(a) 1 PFR (freeboard) & (b) 2 PFRs in series (fluidized bed + freeboard)
(O: Experimental data, SD: San Diego [32], D: Dagaut et al. [29], S: Sung et al. [31],
Q: Qin et al. [30], M: Marinov et al. [28])
ZF (m)
0.00 0.05 0.10 0.15 0.20 0.25
YC
3H
8 (
vo
l%)
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
ZF (m)
0.00 0.05 0.10 0.15 0.20 0.25
YC
3H
8 (
vol%
)
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
(a) M
D
GRI
Q
S
SD
TB = 923 K
MDGRI
Q SSD
TB = 923 K
(b)
62
The discrepancy between the predicted and measured YC3H8 axial profiles could be due to: (1) the
absence of important gas-phase reactions in the models, (2) the production of free radicals and
intermediates upstream of the bed surface and/or (3) gas backmixing in the freeboard.
The possibility that important gas-phase reactions may not be included in the reaction model
motivated the use of six different well established microkinetic schemes. These models contain
state of the art chemistry networks. Furthermore, the models of Dagaut et al. [29] and
Marinov et al. [28] have been validated at conditions (temperature, pressure and composition)
similar to the ones used in the present study.
Since in-bed conversion was observed at the bed temperatures of 853 K and 923 K, free radicals
and intermediates were produced upstream of the freeboard. These species may have been
transported in sufficient amount into the freeboard to accelerate the reaction.
5.2.4.2.2 Accounting for upstream conversion
To model the in-bed conversion, one plug flow reactor was added upstream of the freeboard. It
has been previously shown by Lorences et al. [38] that the gas-phase hydrodynamics in fluidized
beds are very close to plug flow when Ug ≤ 6 × Umf. The mixed-cup propane/air composition at
the sparger tip was used as the inlet boundary conditions and the temperature was assumed
constant at the bed temperature. This first reactor residence time was adjusted to match measured
propane composition at the fluidized bed/freeboard interface. Therefore, the initial condition for
the second reactor modeling the freeboard contained reactive intermediates produced in the first
reactor.
Figure 5.3(b) shows the experimental and predicted YC3H8 obtained with the six microkinetic
mechanisms for a bed temperature of 853 K. With the GRI mechanism, a long induction time and
no propane combustion was obtained within 0.4 m of the bed surface (not shown in
Figure 5.3(b)). On the other hand, with the other five kinetic schemes, the onset of combustion
63
occurred earlier and immediately at the freeboard inlet. Furthermore, the reaction rate of propane
and the total amount of propane consumed in the freeboard increased such that the calculated
YC3H8 axial profiles yielded a better fit of the experimental data. At ZF = 0.31 m, the predicted
YC3H8 ranged from 1.13% (SD) to 0.06% (M) compared to the experimental measurement of
0.07%. The mechanism of Marinov et al. [28] yielded the best fit of the experimental data.
Figure 5-5: Performance of the mechanism of Marinov et al. [28] at low TB (818 K and 830 K)
and with the addition of H2O2
ZF (m)
0.0 0.1 0.2 0.3 0.4
YC
3H
8 (
vo
l%)
0.0
0.4
0.8
1.2
1.6
2.0
818
833
Marinov et al. [28]
ZF (m)
0.0 0.1 0.2 0.3 0.4
YC
3H
8 (
vol%
)
0.0
0.4
0.8
1.2
1.6
2.01 PFR
2 PFRs
1 PFR + 0.54 % H2O
2
Exp. data
Marinov et al. [28]
TB = 853 K
TB (K)
(a)
(b)
YC3H8,MC
= 1.30 %
YC3H8,MC
= 1.95 %
64
The addition of a PFR upstream of the freeboard also accelerated the propane combustion in the
freeboard at a bed temperature of 923 K as shown in Figure 5.4(b). The 6 kinetic mechanisms
predicted an onset of combustion at the bed surface (ZF = 0.0 m) and faster propane combustion
rates in the freeboard compared to Figure 5.4(a), which yielded a better fit between the calculated
and experimental YC3H8 axial profiles. At ZF = 0.16, the predicted YC3H8 ranged between 0.22%
(SD) and 0.00% (GRI) compared to the measured value of 0.02%. The mechanisms of
Marinov et al. [28] (M) and Qin et al. [30] gave the best fit of the experimental data.
The validity of the six microkinetic schemes can also be evaluated by comparing the gas
residence time necessary in the fluidized bed (PFR) model to obtain the correct propane in-bed
conversion with experimental value. The actual average gas residence time inside the fluidized
bed can be estimated from the bed height (0.29 m), the superficial gas velocity at the bed
temperature (0.17 m/s) the void fraction inside the bed. Furthermore, the void fraction can be
calculated from the pressure drop across the bed (2.6 kPa), the solids density (2650 kg/m3s) and
the bed height (0.29 m). The average gas residence time inside the fluidized bed was estimated as
of 1.1 s.
Table 5.1 lists the gas residence time in the fluidized bed (PFR) model that was necessary to
obtain the correct propane conversion at the freeboard inlet for each microkinetic scheme. The
GRI mechanism required gas residence times that were significantly higher than the estimated
value, particularly at TB = 853 K. The other mechanisms required gas residence time that were on
the same order as the experimental value.
The propane freeboard conversion predicted by the mechanism of Marinov et al. [28] (M) also
agreed well with measurements at bed temperatures of 818 K and 830 K. Figure 5.5(a) shows the
experimental and predicted YC3H8 axial profiles at these conditions. At TB = 833 K, partial
propane combustion occurred in the freeboard, which produced a significant temperature
increase, as previously discussed and shown in Figure 5.2. On the other hand, a bed temperature
65
of 818 K did not produce combustion in the freeboard region. Marinov et al. [28] fits the
experimental data well as combustion was predicted at TB = 833 K. Furthermore, very little
propane consumption was predicted at TB = 818 K in agreement with the experiments. These
simulations were performed by taking into account the in-bed conversion in the fluidized bed
region. Furthermore, a fluid bed gas residence time of 1.2 second was used, which corresponds to
the residence time used for the simulation at TB = 853 K (see Table 5.1). The good agreement
between the model of Marinov et al. [28] and the present experimental results may be explained
by: (1) a higher number of species and elementary reactions that better reproduce the reaction and
(2) the fact that the model was validated for conditions (pressure, temperature and mixture
compositions) similar to the ones tested in the present study.
Table 5.1: Average gas residence time in fluidized bed
Microkinetic
mechanism Details
ττττB (s)
TB = 853 K TB = 923 K
Dagaut et al. [29] (D) 50 species
274 elementary reactions 2.3 1.4
GRI v3 [33] (GRI) 53 species
325 elementary reactions 16.1 3.3
Marinov et al. [28] (M) 126 species
638 elementary reactions 1.2 0.6
Sung et al. [31] (S) 92 species
621 elementary reactions 1.9 1.0
San Diego [32] (SD) 46 species
235 elementary reactions 2.9 2.2
Qin et al. [30] (Q) 70 species
463 elementary reactions 2.2 1.0
A set of simulations were performed in order to identify the chemical species and/or free radicals
produced in the fluid bed that were responsible for accelerating combustion in the freeboard:
66
chemical species were removed one-by-one from the inlet of the second PFR modeling the
freeboard. Using this procedure, the reactive intermediates that had no influence on the freeboard
combustion were gradually eliminated. Figure 5.5(b) shows the experimental and predicted axial
profiles of YC3H8 in the freeboard by the model of Marinov et al. [28] for the cases where:
1) Only the freeboard was modeled and in-bed combustion was not accounted for (1 PFR).
2) The fluidized bed was modeled with one PFR and the gas residence time was adjusted to
obtain the correct in-bed propane conversion at the fluidized bed/freeboard interface. The
inlet boundary condition of the second PFR (freeboard) contained all the reactive
intermediates produced by the microkinetic models in the first reactor.
3) Only the freeboard was modeled and the H2O2 produced in the fluidized bed model of
point (2) above was introduced at the freeboard inlet along with the propane/oxygen
mixture.
Figure 5.5(b) shows that hydrogen peroxide produced upstream of the bed surface could account
for most of the increased reaction rate in the freeboard. This was the case for all six microkinetic
models used in this study. Hydrogen peroxide is known to be the main chain branching agent in
the tested temperature range. These results suggest that the production of hydrogen peroxide in
the fluidized bed region may promote combustion reactions in the freeboard for partial in-bed
conversion conditions.
5.2.4.2.3 Gas backmixing in the freeboard
During the induction period, the initial hydrocarbon decomposition produces free radicals and
intermediates that promote the overall reaction. In free radical systems, it is therefore possible
that gas backmixing may lead to lower induction times and higher conversion rates compared to
an ideal plug flow reactor. To characterize the freeboard region, a series of gas tracer experiments
was conducted at ambient temperature over the bed surface at ZF ≥ 0.16 m: the gas-phase
hydrodynamics was found to be plug flow. However, gas backmixing may occur closer to the bed
surface and especially in the transport disengagement height (TDH) where solid particles are
present (0.11 m – as shown in Figure 5.1). To investigate the effect of backmixing on propane
conversion in the TDH, the latter was modeled as one perfectly mixed reactor: a lower
conversion was obtained compared to the case of a plug flow reactor with all microkinetic
67
models. These results suggest that backmixing in the freeboard could not explain the discrepancy
between predicted and measured YC3H8 axial profiles.
5.2.5 Conclusions
Propane combustion experiments were performed in the freeboard region of a fluidized bed of
inert sand particles at low bed temperatures (818 K ≤ TB ≤ 923 K) and low gas superficial
velocity (Ug = 2.1 × Umf). The freeboard region was characterized by simultaneous measurements
of solids flux, chemical composition, temperature and pressure. Experiments showed that
combustion in the freeboard occurred within 0.06 m of the bed surface for TB ≥ 833 K. In-bed
conversion decreased with declining bed temperature and freeboard combustion resulted in a
significant temperature increase. A reaction model was developed to predict propane induction
time and reaction in the freeboard region. Six different gas-phase propane combustion
microkinetic mechanisms were used and the predicted YC3H8 axial profiles were compared to the
experimental measurement – all six mechanisms predicted lower reaction rates compared to the
experimental results. The models also predicted the production of H2O2 during the partial
oxidation and decomposition of propane upstream of the freeboard, which is known as the main
chain branching agent in the considered temperature range. The introduction of hydrogen
peroxide at the inlet of the freeboard region in the simulations significantly accelerated the
propane combustion rates and resulted in a better agreement between predicted and measured
YC3H8 axial profiles. These results suggest that autoignition and gas-phase combustion in the
freeboard region can be promoted by the production of free radicals during reactions in the
fluidized bed region. The model of Marinov et al. [28] yielded the best quantitative agreement
with the experimental results among the considered microkinetic schemes.
5.2.6 Nomenclature
dP average particle size (µm)
GSU solids upward flux or entrainment flux (kg/m2s)
r radial position (m)
68
R reaction radius (0.2 m)
τB average gas residence time in fluidized bed (s)
T temperature (K)
TB fluidized bed temperature (K)
Ug superficial gas velocity (m/s)
Umf minimum fluidization velocity (m/s)
Yi volume fraction of species i (vol%)
Z axial position (m)
ZF axial position in the freeboard (m)
5.2.7 References
[1]Botella P, López Nieto JM, Solsona B. Selective oxidation of propene to acrolein on Mo-Te
mixed oxides catalysts prepared from ammonium telluromolybdates. J Mol Catal A-Chem
2002;184:335-347.
[2]Lorences MJ, Patience GS, Cenni R, Diez F, Coca J. VPO transient lattice oxygen
contribution. Catal Today 2006; 112:45-48.
[3]Malleswara Rao TV, Deo G. Ethane and Propane Oxidation over Supported V2O5/TiO2
Catalysts: Analysis of Kinetic Parameters. Ind Eng Chem Res 2007;46:70-79.
[4]Hutchenson KW, La Marca C, Patience GS, Laviolette J-P, Bockrath RE. Parametric study of
butane oxidation in a circulating fluidized bed reactor. Submitted to Appl Catal A-Gen 2009.
[5]Van der Vaart DR. The chemistry of premixed hydrocarbon/air combustion in a fluidized bed.
Fuel 1988;67:1003-1007.
[6]Van der Vaart DR. Freeboard ignition of premixed hydrocarbon gas in a fluidized bed.
Combust Flame 1988;71:35-39.
[7]Brokaw RS, Jackson JL. Effect of temperature, pressure, and composition on ignition delays
for propane flames. P 5th Combust Inst 1955:563-569.
69
[8]Chang CJ, Thompson AL, Winship RD. Ignition delay of propane in air between 725-880oC
under isothermal conditions. P 7th Combust Inst 1959:431-435.
[9]Burcat A, Lifshitz A. Shock-tube investigation of ignition in propane-oxygen-argon mixtures.
P Combust Inst 1971: 745-755.
[10]Burcat A, Scheller K, Lifshitz A. Shock-tube investigation of comparative ignition delay
times for C1-C5 alkanes. Combust Flame 1971;16:29-33.
[11]Cathonnet M, Boettner JC, James H. Experimental study and numerical modeling of high
temperature oxidation of propane and n-butane. P 18th Combust Inst 1981:903-913.
[12]Kin K, Soo Shin K. Shock tube and modeling study of the ignition of propane. Bull Korean
Chem Soc 2001;22(3):303-307.
[13]Penyazkov OG, Ragotner KA, Dean AJ, Varatharajan B. Autoignition of propane-air
mixtures behind reflected shock waves. P Combust Inst 2005;30:1941-1947.
[14]Horning DC, Davidson DF, Hanson RK. Study of the high-temperature autoignition of n-
alkane/O2/Ar mixtures. J Propul Power 2002;18(2):363-371.
[15]Zhukov VP, Sechenov VA, Starikovskii. Autoignition of a lean propane-air mixture at high
pressures. Kinet Catal 2005;46(3):319-327.
[16]Freeman G, Lefebvre AH. Spontaneous ignition characteristics of gaseous hydrocarbon-air
mixtures. Combust Flame 1984;58:153-162.
[17]Cadman P, Thomas GO, Butler P. The auto-ignition of propane at intermediate temperature
and high pressures. Phys Chem Chem Phys 2000;2:5411-5419.
[18]Hesketh RP, Davidson JF. Combustion of methane and propane in an incipiently fluidized
bed. Combust Flame 1991;85:449-467.
[19]Wen CY, Chen LH. Fluidized bed freeboard phenomena: entrainment and elutriation. AIChE
J 1982;28(1):117-128.
[20]Kunii D, Levenspiel O. Entrainment of solids from fluidized beds I.Hold-up of solids in the
freeboard II.Operation of fast fluidized bed. Powder Technol 1990;61:193-206.
70
[21]De Lasa HI, Grace JR. The influence of the freeboard region in a fluidized bed catalytic
cracking regenerator. AIChE J 1979;25(6):984-991.
[22]Chen LH, Wen CY. Model of solid gas reaction phenomena in the fluidized bed freeboard.
AiChEJ 1982;28(6):1019-1027.
[23]Walsh PM, Chaung TZ, Dutta A, Beér JM, Srofim AF. Particle entrainment and nitric oxide
reduction in the freeboard of a fluidized coal combustor. ACS Fuels Volumes 1982;27(1):243-
261.
[24]Sotudeh-Gharebagh R, Mostoufi N. Simulation of a catalytic turbulent fluidized bed reactor
using the sequential modular approach. Fuel Process Technol 2003;85:189-200.
[25]Dounit S, Hemati M, Andreux R. Modeling and experimental validation of a fluidized-bed
reactor freeboard region: application to natural gas combustion. Chem Eng J 2008;140:457-465.
[26]Hartman M, Trnk O, Pohořelý, Svoboda K. High-temperature reaction in the freeboard
region above a bubbling fluidized bed. Ind Eng Chem Res 2010;XX:XX.
[27]Buda F, Glaude PA, Battin-Leclerc F, Porter R, Hughes KJ, Griffiths JF. Use of detailed
kinetic mechanisms for the prediction of autoignitions. J Loss Prevent Proc 2006;19:227-232.
[28]Marinov NM, Pitz WJ, Westbrook CK, Hori M, Matsunaga N. An experimental and kinetic
calculation of the promotion effect of hydrocarbons on the NO-NO2 conversion in a flow reactor.
P Combust Inst 1998;27: 389-396.
[29]Dagaut P, Cathonnet M, Boettner JC, Gaillard F. Kinetic oxidation of propane oxidation.
Combust Sci Technol 1987;56:23-63.
[30]Qin Z, Lissianski VV, Yang H, Gardiner Jr WC, Davis SG, Wang H, Combustion chemistry
of propane: a case study of detailed reaction mechanism optimization. P Combust Inst 2000;
28:1663–1669.
[31]Sung CJ, Li B, Law CK, Wang H. Structure and sooting limits in counterflow methane/air
and propane/air diffusion flames from 1 to 5 atmospheres, 27th Symp (Intl) Combust. 1998:1523-
1530.
71
[32]Chemical-Kinetic Mechanisms for Combustion Applications, San Diego Mechanism web
page, Mechanical and Aerospace Engineering (Combustion Research), University of California at
San Diego. (http://maeweb.ucsd.edu/~combustion/cermech/)
[33]Smith GP, Golden DM, Frenklach M, Moriarty NW, Eiteneer B, Goldenberg M, Bowman
CT, Hanson RK, Song S, Gardiner Jr WC, Lissianski VV, Qin
Z. http://www.me.berkeley.edu/gri_mech/
[34]Jeng R-H, Altwicker ER, Morgan III MM, Wilk RD. Propane combustion in a spouted bed
combustor II: a modeling study in the bed region. Combust Sci Technol 2001;170:119-149.
[35]Sotudeh-Gharebagh R, Chaouki J. The heterogeneous and homogeneous combustion of
methane over inert particles. Can J Chem Eng 2003;81:1182-1191.
[36]Ranzi E, Faravelli T, Gaffuri P, Pennati GC, Sogaro A. A wide range modeling study of
propane and n-butane oxidation. Combust Sci Technol 1994;100:299-330.
[37]Wilk RD, Cernansky NP, Cohen RS. The oxidation of propane at low and transition
temperatures. Combust Sci Technol 1986;49:41-78.
[38]Lorences M, Laviolette J-P. Fluid bad gas RTD: effect of fines and internals. Powder
Technol 2006; 168 (1): 1-9.
72
CHAPITRE 6 NOUVELLE MÉTHODE DE SPECTROSCOPIE POUR
LA MESURE SIMULTANÉE ET QUANTITATIVE DE LA
COMPOSITION CHIMIQUE DE LA PHASE GAZEUSE ET DE LA
FRACTION DE SOLIDES
Cet article a été accepté par AIChE journal en 2010. L’article est présentement publié en ligne.
Un brevet sur la méthode de mesure décrite dans cet article a été déposé aux États-Unis. La
confirmation du dépôt du brevet est présentée à l’annexe 9.
6.1 PRÉSENTATION DE L’ARTICLE
L’objectif de cet article est de développer une nouvelle méthode spectroscopique pour la mesure
simultanée et quantitative de la fraction de solides et de la composition chimique de la phase
gazeuse. La méthode est utilisée avec une sonde à fibres optiques infrarouges afin d’effectuer des
mesures in situ et en ligne dans des systèmes gaz/solides. Les résultats démontrent que les valeurs
d’absorbance résultantes de la fraction de solides et de la composition chimique sont additives.
Des mesures instantanées sont effectuées à une fréquence atteignant 4.5 Hz avec une précision
considérable (erreur relative maximale d’environ 5% sur les mesures moyennées dans le temps).
Des mesures sont effectuées dans un lit fluidisé afin de mesurer la fraction volumétrique d’un
traceur gazeux dans la phase bulle et la phase émulsion.
73
SIMULTANEOUS QUANTITATIVE MEASUREMENT OF GASEOUS SPECIES
COMPOSITION AND SOLIDS VOLUME FRACTION IN A GAS/SOLID FLOW
Jean-Philippe Laviolette, Gregory S. Patience and Jamal Chaouki
(Published in AIChE Journal, 2010)
6.2 SIMULTANEOUS QUANTITATIVE MEASUREMENT OF GASEOUS
SPECIES COMPOSITION AND SOLIDS VOLUME FRACTION IN A
GAS/SOLID FLOW
A novel spectroscopic method was developed to measure quantitatively and simultaneously
solids volume fraction (1-ε) and gaseous species composition (Yi) in a gas/solid system. The
method was comprised of an FT-IR coupled to a fibre-optic probe that could perform real-time
and in-situ measurements of absorbance. The effect of (1-ε) and Yi on the absorbance spectra
were additive and could be independently calibrated. Experiments were conducted with
alkane/nitrogen mixtures and two types of particles: sand and FCC. Fuel mole fractions and (1-ε)
were varied between 1.8 - 10.1 mol% and 0 - 0.45, respectively. The relative errors for Yi time-
averaged measurements were below 6% and the error increased significantly with decreasing
beam intensity. A proof of concept for a novel application in fluidized beds was also completed:
the fibre-optic probe was used to measure the molar fraction of a tracer gas inside the emulsion
and bubble phases during gas tracer experiments.
6.2.1 Introduction
Multiphase processes are common to many industries: petrochemical, chemical, pharmaceutical,
energy, food, pulp and paper, etc. They are also present throughout the process chain: energy
generation (combustors), transformation (gasifiers, chemical reactors, mixers and dryers),
purification/separation (cyclones and precipitators) and waste disposal (chimneys). System
characterization through the measurement of species composition and solids volume fraction is
critical for process control, optimization and trouble-shooting in the industries and laboratories. It
is also essential for the development of hydrodynamic (gas/liquid/solid) and kinetic models.
74
Chemical species composition and solid volume fraction are currently measured independently
with two different techniques.
Several non-invasive and invasive techniques have been developed to measure solids volume
fraction in two- and three-phase systems. Invasive techniques involve the insertion of probes to
measure the local capacitance [1, 2, 3], conductivity [4, 5, 6], fluorescence [7], forward-scattering
[8] or back-scattering of light in the medium [2, 9, 10, 11]. The solids volume fraction is derived
from the measurement, which can be made in-situ and in real-time. In liquid/solid or
gas/liquid/solid systems, the slurry phase can also be sampled with special probes to measure the
liquid and solids volume fraction [7, 12]. Non-invasive techniques include tomography [13, 14,
15, 16], densitometry [11] and the analysis of ultrasonic wave distortions [17]. These techniques
produce time- or space-averaged maps solids volume fraction in multiphase systems with varying
resolution.
Species concentrations in the gas or liquid phases can be analyzed continuously or intermittently
[12] with analytical instruments (chromatograph, spectrometer, etc.) via sampling. Furthermore,
in the context of the U.S. Food and Drug Administration (FDA) process analytical technology
(PAT) initiative, several publications have reported the development of infrared applications for
the measurement of chemical species composition in multiphase samples [18, 19]. Similar
applications have also been developed for the in-situ monitoring of reactions at the surface of
catalysts [20, 21]. These systems function in transmission and diffuse reflectance modes (on the
solid surface) and can measure species composition the continuous phases (liquid and gas) as
well as characterizing qualitatively the chemical composition of the discrete phase (solid particles
and liquid droplets) [22, 23, 24, 25, 26]. All these techniques measure species concentrations in
the continuous phase as a spatial average between the different regions of varying solids volume
fractions (bubble and emulsion phases in fluidized bed, for example).
A technique that measures species composition and solids volume fraction simultaneously yields
much more information into multiphase processes since solids volume fraction and species
concentrations are dependent parameters. These two parameters are coupled through the reaction
75
kinetics and the hydrodynamics. During reaction, solids volume fraction influences chemical
reactions through heterogeneous catalytic and/or inhibitive effects as well as through the thermal
balance. Furthermore, at the hydrodynamic level, multiphase systems are often characterized by
different regions of varying void fraction and the species composition in these regions are
fundamental to the process characterization. For example, bubbling and turbulent gas/solid
fluidized beds are characterized by a dense phase (emulsion) and a dilute phase (bubbles).
Bubbles cause gas by-passing that lowers the process yield. The simultaneous measurement of
solids volume fraction and gaseous species composition discriminates between the reactants in
the dilute and dense phases and this measurement can hardly be made with existing methods.
Infrared (IR) spectroscopy offers a great potential for the simultaneous measurement of solids
volume fraction and gaseous species composition. IR spectroscopic applications in multiphase
systems are affected by the solid phase characteristics. The solids volume fraction and particle
size have been shown to influence the interaction between the light and the solids, which affect
the path length of the IR beam and the effective sample size [20, 23]. An increase in particle size
has been reported to decrease the fraction of scattered light and to increase penetration depth of
light into the solid particles, resulting in increased absorbance and an upward shift of the
absorbance baseline in diffuse reflectance spectroscopy [23]. Therefore, current IR spectroscopic
applications in multiphase system rely on the uniformity of the solid samples with time (between
measurements). This is achieved by performing the spectral acquisition in immobile samples
such as solid tablets [25, 27, 28, 29, 30]. On the other hand, measurements in dynamic systems
(fluidized beds or mixers, for example) require shutdown prior to the spectral acquisition [31,
32], the synchronization of the spectral acquisition to a specific vessel position [33] or
measurement in regions of high solids density [34]. Non-intrusive NIR monitoring of powder
blend homogeneity has also been conducted in symmetrical mixers to ensure the temporal
uniformity of the solids volume fraction when complete mixing is reached [35]. Finally,
mathematical treatments of the spectra, such as the derivative treatment, are often used to
minimize the effects of baseline shifts due to changes in the solid phase properties [23]. However,
multiphase systems are generally characterized by solids volume fractions that are heterogeneous
in space and time. Therefore, the in-situ and real-time measurement of species concentrations in
the different regions of a multiphase system requires the simultaneous determination of solids
76
volume fraction. The measurement volume can be made independent of the solid properties by
inclining the emitting and receiving fibres to form a convergent scheme [36]. This principle can
be applied to an IR fibre-optic probe to measure species concentrations.
In multiphase systems, the movement of powders makes spectroscopy measurements and
interpretation more complex. The measured absorbance spectra are influenced by variations in
the solids volume fraction and the average distance from the sample to the probe. When samples
are moving, different parts of the spectrum may also be influenced to a varying degree by local
heterogeneities (different material components). The effects of moving particles on Fourier-
Transform spectroscopy have been investigated by De Paepe et al. [37], Berntsson et al. [38] and
Andersson et al. [39]. It was shown that the movement of the particles produced artefacts on the
measured spectrum at certain wavenumbers that depended on the timescale of the spectral scan
compared to the rate of movement of the particles. The wavenumber at which the artefact
appeared increased with increasing modulation frequencies such that the effects of moving
particles could be completely eliminated at the wavenumbers of interest with a sufficiently high
modulation frequency.
In the present study, a novel spectroscopic method was developed for the simultaneous and
quantitative measurement of gaseous species composition (Yi) and solids volume fraction (1-ε).
Experiments were conducted using near- and mid-IR Fourier transform transmission
spectroscopy (6000-1000 cm-1). A fibre-optic probe was constructed to perform in-situ and real-
time measurements in a gas/solid flow of methane/nitrogen and silica sand particles
(dp = 290 µm). Absorbance spectra were recorded at a frequency of 4.5 Hz for a period of 75
seconds and both Yi and (1-ε) were evaluated from each spectrum. Methane molar fractions in
nitrogen were measured over a range of 0-10.1 vol%. Furthermore, the solids volume fraction
was varied between 0-0.1. The fibre-optic probe was also used inside a fluidized bed of FCC
catalyst during gas tracer experiments to measure the tracer (CH4) molar fraction inside the
bubble and emulsion phases ((1-ε) = 0.45).
77
6.2.2 Experimental
Experiments were conducted with an FT-IR (Varian Excalibur series 3100) equipped with a high
sensitivity mercury-cadmium-telluride (MCT) detector cooled with liquid nitrogen. In a first
series of tests, the FT-IR was used with a gas flow cell located in its sample compartment as
shown in Figure 6.1. The IR beam coming from the interferometer was transmitted through the
gas flow cell and the absorbance spectrum was measured by the MCT detector. Prior to each
experiment, a background spectrum was measured by flushing the gas flow cell with air and by
co-adding 20 spectra. Then, a gas/solid system was simulated in the FT-IR sample compartment
by injecting a mixture containing 1.8 vol% C3H8 in air inside the gas flow cell and by placing a
metal mesh (open area ratio of 16% and thickness of 1 mm) in the path of the IR beam.
Figure 6-1: FT-IR apparatus
A second series of experiments were performed with a mid-IR fibre-optic probe. In-situ
measurements of gaseous species composition and solids volume fraction were conducted inside
a gas/solid flow of CH4/N2 mixtures and silica sand particles (dp = 290 µm). A schematic of the
apparatus is shown in Figure 2. The probe was composed of two parallel fluoride glass fibre-
optics (an emitting and a receiving fibre-optic) with a numerical aperture of 0.2 and a core
diameter of 600 µm. The fibre-optic probe was connected to the FT-IR with a Harrick
Fibremate™. A planar gold-coated mirror was positioned perpendicularly to the probe tip at a
distance of 5 mm. The mirror reflected the emitted IR beam to the receiving fibre-optic, which
resulted in a measurement volume between the probe tip and the mirror as shown in grey on
Figure 6.2. Experiments were conducted by first measuring a background spectrum - a CH4/N2
MCT detector
FT-IR interferometer
Metal mesh
Gas flow cell
(C3H8/Air)
78
mixture was injected inside the measurement and a background spectrum was obtained from the
co-addition of 20 spectra. Then, the measurement volume was flushed with air and the funnel was
filled with sand particles to initiate a flow of solids. Absorbance spectra were measured at a
frequency of 4.5 Hz (temporal resolution of 0.22 s) and the methane molar fraction was measured
simultaneously to the solids volume fraction. The flow of solid particles was varied by changing
the diameter of the funnel diameter at its throat. The corresponding solids volume fraction for
each throat diameter was measured with a conventional back-scattering fibre-optic probe between
the IR probe and the mirror. Four solids volume fractions were used: 0, 0.015, 0.038 and 0.097.
Furthermore, four different CH4/N2 compositions were injected from calibrated gas cylinder in
the measurement volume: 0, 0.1, 3.25 and 10.1 vol%.
Figure 6-2: Fibre-optic probe apparatus
Gas tracer experiments were also conducted in a small scale fluidized bed reactor (I.D. = 5 cm)
illustrated in Figure 6.3 and the fibre-optic probe was used to measure the tracer molar fraction
inside the bubble and emulsion phases. The reactor was filled with FCC particles (dp = 83 µm
and 14% fines (dp ≤ 44 µm)) and was fluidized with nitrogen at a superficial gas velocity of
2.6 mm/s (Umf = 2.5 mm/s and Umb = 2.7 mm/s). The expanded bed height was 12.5 cm and the
fibre-optic probe was inserted in the bed at a height of 7 cm. A mirror was positioned
Emitting fiber-optic
Receiving fibre-optic FT
-IR
Planar gold-coated mirror
Funnel filled with sand particles
t = 5 mm
79
perpendicularly at the probe tip and gas bubbles were produced in the fibre-optic probe
measurement volume by injecting a mixture containing 10.1% CH4 + 89.9% N2 through a
downward facing sparger. The CH4/N2 mixture was injected at a flow rate of 10 mL/s by
manually opening a toggle valve for roughly 0.5 s at an interval of approximately 11 seconds.
Figure 6-3: Fluidized bed reactor
6.2.3 Results and discussion
During transmission spectroscopy in a gas/solid sample, both solid particles and gaseous species
contribute to the absorbance spectrum according to equation (6.1): the total absorbance at
wavelength λ has a contribution from the solids volume fraction and from the gaseous species
composition. Solid particles reduce the incident beam intensity by absorption, reflection and
diffusion. Solids can also influence the path length and sample volume such that the absorbance
due to the chemical species is a function of solids volume fraction.
( ) ( )[ ]ελλελ −+= − 11 fnAAA ,Y,,Total i (6.1)
In the specific case of transmission spectroscopy where the effect of solids fraction on path length
and sample volume can be neglected, equation (6.1) can be simplified to:
CH4/N2
N2
IR fibre optic probe
FT
-IR
Gold coated
mirror Bubble Sand
Porous plate
80
( ) λλελ ,,1, iYTotal AAA += − (6.2)
Equation (6.2) is valid when the reflected or diffused beam intensity measured is negligible
compared to the transmitted light intensity. In equation (6.2), the effects of solids volume fraction
and gaseous species composition on the absorbance spectrum are independent. Therefore, the
effect of (1-ε) and Yi can be calibrated independently.
6.2.3.1 Experiments with the FT-IR
A first series of experiments were conducted with the FT-IR of Figure 6.1. A background
spectrum of air was taken and 4 absorbance spectra were subsequently measured for the
following conditions:
• Air and (1-ε) = 0 (Figure 6.4a)
• Air and (1-ε) = 0.84 (Figure 6.4b)
• 1.8% C3H8 + 98.2% air and (1-ε) = 0 (Figure 6.4c)
• 1.8% C3H8 + 98.2% air and (1-ε) = 0.84 (Figure 6.4d)
Figure 6.4a shows the absorbance spectrum of air, which corresponded to the background
absorbance spectrum and resulted in a baseline of zero absorbance. Two calibration absorbance
spectra were measured to independently evaluate the effect of the propane molar fraction and the
solids volume fraction. The absorbance spectrum for air and a solids volume fraction of 0.84 is
shown in Figure 6.4b - the baseline absorbance increased almost uniformly across the spectrum
from to 0 to a higher absorbance. The average baseline absorbance was 0.77 in the region of
3436.90-3427.98 cm-1. This specific range of wavenumbers (3436.90-3427.98 cm-1) was chosen
to evaluate the average baseline absorbance since it was characterized by a good signal-to-noise
ratio and since it was not absorbed by propane. The second calibration spectrum was obtained by
81
injecting a mixture of 1.8% C3H8 in air inside the gas flow cell and by removing the metal mesh
from the sample compartment. The injection of C3H8 resulted in the formation of two absorbance
peaks at 2967.96 and 1471.68 cm-1, as illustrated in Figure 6.4c. The average peak height of the
tallest peak (at 2967.96 cm-1) was measured as 0.92 in the region of 2969.65-2966.27 cm-1.
Figure 6-4: Simultaneous measurement of YC3H8 and (1-ε)
Finally, the propane concentration and the solids volume fraction were measured simultaneously.
The mixture of 1.8% C3H8 was injected inside the gas flow cell and the metal mesh was inserted
inside the sample compartment. The resulting absorbance spectrum corresponded to the addition
of both calibration spectra as observed in Figure 6.4d: the baseline absorbance was 0.79
(compared to 0.77) and the average C3H8 peak height above the baseline was 0.95 (compared to
Wavenumber (cm-1
)
100020003000400050006000
A-0.3
0.0
0.3
0.6
0.9
1.2
1.5
1.8
2.1
Wavenumber (cm-1
)
100020003000400050006000
A
-0.3
0.0
0.3
0.6
0.9
1.2
1.5
1.8
2.1
Wavenumber (cm-1
)
100020003000400050006000
A
-0.3
0.0
0.3
0.6
0.9
1.2
1.5
1.8
2.1
AC3H8 = 0.92
A(1-ε) = 0.77
Wavenumber (cm-1
)
100020003000400050006000
A
-0.3
0.0
0.3
0.6
0.9
1.2
1.5
1.8
2.1
A(1-ε) = 0.79
AC3H8 = 0.95
(a) (b)
(d)(c)
Baseline (air) (1-ε) = 0.84
YC3H8 = 1.8%
(1-ε) = 0.84
YC3H8 = 1.8%
82
0.92). Therefore, the propane concentration and solids volume fraction were measured
simultaneously with a relative error of approximately 3% if a linear relation between absorbance
and the two parameters is assumed. Equation (6.2) was valid in that particular case since the
metal mesh had no appreciable influence on the path length of the IR beam – the path length was
defined by the width of the sample compartment.
6.2.3.2 Experiments with the fibre-optic probe
A fibre-optic probe was built to conduct in-situ measurements of solids volume fraction and
species concentrations inside a flow of silica sand particles. Using the fibre-optic probe resulted
in a significant loss in IR beam intensity, which occurred most likely at the probe tip. However,
some signal loss probably occurred at the FT-IR/probe interface (Harrick Fibremate™) as well.
Designing a system to yield a sufficiently intense transmitted IR beam was a challenge. The first
fibre-optic probe prototype was built with two fibre-optics with a core diameter of 300 µm and no
IR signal was ever measured by the FT-IR. The feasibility of using a more powerful external IR
source was examined, but the coupling of the IR beam with the FT-IR interferometer proved to
be particularly complex and yielded a significant risk of reducing the IR beam intensity below
that of the internal FT-IR source. The installation of an external source was therefore abandoned.
The construction of a new probe with core diameters of 600 µm and a complete re-alignment of
the FT-IR increased the transmitted IR beam intensity sufficiently to perform the current
experiments.
Another means of increasing the IR signal intensity consisted of installing a more sensitive
detector. Indium antimonide (InSb) detectors can offer higher signal-to-noise ratios compared to
MCT detectors at the wavenumbers of interest in the present study and an InSb detector was
therefore tested. However, the selected InSb detector functioned at relatively low modulation
frequencies (5-10 kHz) compared to the MCT detector (5-80 kHz) and this proved to be
problematic in gas/solid systems where particles are moving. Particle movement produced
artefacts (noise) in the interferograms and absorbance spectra and an initial analysis was
performed to eliminate the artefacts from the wavenumbers of interest.
83
Figure 6-5: Effect of particle movement:
(a) typical interferograms for moving particles [(1-ε ) = 0.015];
(b) corresponding single-beam spectra
6.2.3.2.1 Effect of particle movement
Figure 6.5(a) shows typical interferograms measured with the fibre-optic probe in air at different
modulation frequencies and two solids volume fractions. In the absence of solid particles, the
interferogram was characterized by a centreburst at an optical retardation of 0 cm and a signal
decay at higher retardation. With a flow of solids in the probe measurement volume that
corresponded to a time-average solids volume fraction (1-ε ) of 0.015, fluctuations in the signal
appeared. The number of fluctuations decreased as the modulation frequency (f) was increased
from 10 to 80 kHz. Figure 6.5(b) shows the corresponding single-beam spectra. In the absence of
solid particles, the spectrum was characterized by a reasonably high signal-to-noise ratio (S/N)
over the entire spectrum. However, with a flow of solids and (1-ε ) = 0.015, noise was observed
-0.3
0.0
0.3
0.0
Vo
lts
-0.3
0.0
0.3
0.0
Optical retardation (cm)
0.0 0.1 0.2 0.3 0.4
-0.3
0.0
0.3
0.0
0.3
0.6
0.0
0.3
Re
sp
on
se
0.0
0.3
0.0
0.3
Wavenumber (cm-1)
0 1500 3000 4500 6000 7500
0.0
0.3
f = 80 kHz
f = 40 kHz
f = 20 kHz
f = 10 kHz
f = 80 kHz
no solids(a) (b)
0.015)(1 =− ε
0.015)(1 =− ε
0.015)(1 =− ε
0.015)(1 =− ε
f = 80 kHz
f = 80 kHz
f = 40 kHz
f = 20 kHz
f = 10 kHz
no solids
0.015)(1 =− ε
0.015)(1 =− ε
0.015)(1 =− ε
0.015)(1 =− ε
84
in the regions of low wavenumber (high wavelength and low frequency). The noise was produced
by temporal variations in transmitted IR beam intensity due to heterogeneities in the gas/solid
flow. As the modulation frequency was increased from 10 to 80 kHz, the noise moved to lower
wavenumbers (higher wavelengths and lower frequencies). The modulation frequency
corresponds to the number of displacements of 632.8 nm (equal to the laser wavelength) the
moving mirror makes in 1 second in the interferometer. As the mirror speed was increased, a
growing number of wavenumbers were modulated to frequencies that were significantly higher
than the frequency at which the moving silica particles changed the transmitted IR beam
intensity. At high wavenumbers (high frequency), the signal amplitude remained largely
unaffected by the movement of particles. At a modulation frequency of 80 kHz, the artefacts were
outside the wavenumber range of interest (2700-3300 cm-1) and all experiments were therefore
performed with this setting. Tests were also performed at higher solids volume fractions with no
significant effect on the location of the fluctuations observed in the spectra.
6.2.3.2.2 Simultaneous measurement of methane concentration and solids volume fraction
The fibre-optic probe was used to measure in-situ absorbance spectra across the flow of sand
particles. IR spectroscopy is based on the ratio of the background and the measured spectra.
Experimentally, it was simpler to measure a background spectrum with a mixture of CH4/N2
(without solid particles) and conduct the experiments with a flow of particles in air such that
negative CH4 absorbance peaks were obtained. However, the methane molar fractions are
reported as positive values in the present document. The results were unaffected by this
procedure since the IR path length and measurement volume were independent of the solids
volume fraction. When the mirror was removed from the tip of the probe, the signal amplitude
measured by the MCT detector was zero. Therefore, the IR beam collected by the receiving fibre-
optic did not have any contribution from diffuse reflectance on the solids surface and the probe
measurement volume was only defined by the mirror position. Furthermore, the effects of
methane concentration and solids volume fraction on the absorbance spectrum were independent.
A calibration was first performed with the fibre-optic probe to evaluate the effect of methane
molar fraction and solids volume fraction on the absorbance spectrum. The effect of methane
85
molar fraction (YCH4) was calibrated with four mixtures containing 0, 0.1, 3.25 and 10.1 vol% of
methane in nitrogen. The apex of the methane peak (ACH4) was located at a wavenumber of
3017.63 cm-1 and its average absorbance was measured in the region of 3018.96-3016.86 cm-1.
The average absorbance in the region of 3018.96-3016.86 cm-1 is shown as a function of methane
molar fraction in Figure 6.6(a). The relationship between absorbance and YCH4 was non-linear,
but such a deviation from the Beer Law is common in infrared spectroscopy [40].
Figure 6-6: Probe calibration
The effect of solids volume fraction was calibrated by varying the solids flow in the fibre-optic
measurement volume. Four solids volume fractions were used (0, 0.015, 0.038 and 0.097) and the
average baseline absorbance (A(1-ε)) was measured in the regions of 2997.73-2992.04, 3036.70-
3030.72 and 3045.87-3040.39 cm-1 during 75 seconds at a frequency of 4.5 Hz. This range of
wavenumbers was chosen to evaluate the average baseline absorbance since it was characterized
by a good signal-to-noise ratio and since it was not absorbed by methane. Figure 6.7 shows the
history of the baseline absorbance for three solid flows that corresponded to time-averaged solids
volume fractions of 0, 0.015 and 0.097. The measured absorbance fluctuated due to
heterogeneities in the flow of solids. The average baseline absorbance is shown in Figure 6.6(b)
as a function of solids volume fraction and the relationship is close to linear. Cutolo et al. [8]
previously reported a linear relationship between absorbance and solids volume fractions. In the
present case, solids volume fraction was measured by a conventional back-scattering fibre-optic
ACH4
0.0 0.1 0.2 0.3 0.4 0.5 0.6
YC
H4
(vo
l%)
0
2
4
6
8
10
12
ABaseline
0.0 0.2 0.4 0.6 0.8 1.0
(1-ε
)
0.00
0.02
0.04
0.06
0.08
0.10
(a) (b)
R2 = 0.9999 R
2 = 0.9814
( ) BA09579.01 =− ε23.25
8.504.0
CH4
CH4CH4
A
AY
+
+=
86
probe with a different measurement volume compared to the IR fibre-optic probe. Since the
solids volume fraction was not completely constant across the flow of sand particles produced by
the funnel, this may explain the observed deviation from the linear relationship.
Figure 6-7: Solids volume fraction measurement
Finally, the methane concentration and the solids volume fraction were simultaneously measured
in the system. A background spectrum was first taken with a mixture containing 3.25% CH4 in
nitrogen. Then, a flow of sand particles that corresponded to a time-averaged solids volume
fraction of 0.015 was initiated in air. Absorbance spectra were measured in real-time for
75 seconds and at a frequency of 4.5 Hz (temporal resolution of 0.22 second). Figure 6.8(a)
shows the average baseline absorbance (A(1-ε)) and average methane peak absorbance (ACH4) as a
function of time for run #2 (Table 1). Both A(1-ε) and ACH4 fluctuated with time around time-
averaged values of 0.26 and 0.19, respectively. The calibrations (Figure 6.6) were used to
calculate the instantaneous methane molar fraction (YCH4) and solids volume fraction (1-ε) at each
time interval. Figure 6.8(b) shows the calculated YCH4 and (1-ε) as a function of time: both
measurements fluctuated, but the time-averaged values were very close to the real values: 3.24%
(compared to the real value of 3.25%) and 0.019 (compared to the real value of 0.015).
Time (s)
0 15 30 45 60 75
A
0.0
0.2
0.4
0.6
0.8
1.0
1.2
( ) 0.96=−ε1A
( ) 0.24=−ε1A
( ) 0.00=−ε1A
87
Figure 6-8: Simultaneous measurement of (1-ε) and YCH4
Figure 6.9 shows the measured methane molar fraction as a function of time and four horizontal
lines represent the 5% and 10% relative deviation from the real value (3.25% CH4). A majority of
data points (55%) were within 5% of the real value. Furthermore, 87% of data points had a
relative error of 10% or less.
Table 6.1 lists the experiments where solids volume fraction and methane concentration were
measured simultaneously. The methane molar fractions and solids volume fractions fed to the
system are listed for each run as well as the measured time-averaged values obtained from the
Time (s)
0 15 30 45 60 75
A(1
-ε)
0.0
0.5
1.0
1.5
2.0
AC
H4
0.00
0.07
0.14
0.21
0.28
0.35
Time (s)
0 15 30 45 60 75
(1-ε
)
0.00
0.05
0.10
0.15
0.20
YC
H4
(vo
l%)
0.0
0.9
1.8
2.7
3.6
4.5
(a)
(b)
YCH4 = 3.25%
YCH4 = 3.25%
0.015)(1 =− ε
0.015)(1 =− ε
88
absorbance spectra. The measured time-averaged solids volume fraction (1-ε ) and methane
molar fraction ( CH4Y ) were all very close to the real values: the maximum relative error on the
measured CH4Y was 5.8% (run #9). Table 6.1 also shows the percentage of data points that had a
relative error in the measured CH4Y of 5% and 10% or less – this percentage decreased with
increasing solids volume fraction. Increasing the solids volume fraction decreased the accuracy of
the real-time and time-averaged YCH4 measurements.
Figure 6-9: Error on instantaneous YCH4 measurement
6.2.3.2.3 Effects of the transmitted IR beam intensity on the Yi measurement error
As the solids volume fraction increased, the intensity of the transmitted IR beam decreased and
the signal-to-noise ratio (S/N) of the FT-IR decreased accordingly. Figure 6.10a was obtained by
taking a background spectrum of a mixture of 3.25% CH4 in nitrogen and measuring the
absorbance spectra of air with two time-averaged solids volume fractions (0 and 0.015).
Figure 6.10a shows that a time-averaged solids volume fraction of 0.015 resulted in a
significantly lower S/N and the fluctuations of the baseline were of the same order as the methane
peak (located at 3017.63 cm-1).
Time (s)
0 15 30 45 60 75
YC
H4 (
vo
l%)
0
1
2
3
4
5
5%
10%
Error
YCH4 = 3.25%
0.015)(1 =− ε
89
The intensity of the IR beam is a concern with fibre-optic probes since their use significantly
reduces the effective IR signal. Figure 6.10b shows the single beam spectrum of the FT-IR and
the fibre-optic probe for air and (1-ε) = 0: the IR signal intensity was reduced by a factor of
approximately 288 when the fibre-optic probe was installed. Note that the sensitivity of the MCT
detector had to be increased by a factor of 16 with the fibre-optic probe.
Table 6.1: Simultaneous measurement of YCH4 and (1-ε)
Run
Fed to system Measurements
CH4Y
(%)
( )ε−1 CH4Y
(%)
( )ε−1 Percentage (%) of YCH4 measurements within a relative error of
± 5% ± 10%
1 3.25 0.015 3.31 0.019 54 83
2 3.25 0.015 3.24 0.019 55 87
3 10.1 0.015 10.09 0.02 83 98
4 10.1 0.015 10.20 0.018 82 98
5 3.25 0.038 3.31 0.039 38 70
6 3.25 0.038 3.41 0.037 37 63
7 10.1 0.038 10.05 0.041 61 86
8 10.1 0.038 9.95 0.04 63 87
9 3.25 0.097 3.44 0.097 18 31
10 10.1 0.097 10.06 0.098 30 55
For each experimental run listed in Table 6.1, the variance (σ) of the methane molar fraction
measurements was calculated. Figure 6.11 shows the variance as a function of time-average
solids volume fraction and methane molar fraction: increasing the solids volume fraction greatly
increased the variance. This strongly suggests that the accuracy of the fibre-optic probe
90
measurements can be improved by increasing the IR beam intensity. This could be achieved by
using a fibre-optic bundle, optimizing the signal transmission at the probe tip and at the
FT-IR/fibre-optic probe interface (Harrick Fibremate™). The effect of decreasing S/N could also
be minimized by improving the curve smoothing technique to evaluate the baseline absorbance.
Figure 6-10: Sources of error on YCH4 measurement
6.2.3.2.4 Real-time and time-averaged Yi measurements
Real-time YCH4 measurements were characterized by greater errors than the time-averaged values.
For example, Figure 6.9 shows instantaneous YCH4 measurements with relative errors of more
than 10% while the error on the time-averaged YCH4 measurement (Table 6.1 - run #2) was less
than 1%. Furthermore, the maximum relative error on the time-averaged YCH4 measurements was
5.8% (run #9). Applied to a multiphase reactor, the present method could measure precisely the
time-averaged chemical species composition inside the regions of relatively low solids volume
fractions. The technique can also be more sensitive to lower (1-ε) and Yi measurements by
increasing the IR beam path length (t). Since an average species composition between the dense
and dilute phases can be measured by using a sampling probe and analysing the samples, the
chemical composition in the dense phase could be calculated by subtracting the dilute phase
composition from the average (dilute and dense phases) composition.
Wavenumber (cm-1
)
27002850300031503300
A
-0.30
-0.15
0.00
0.15
0.30
0.45
0.60
0.75
Wavenumber (cm-1
)
02000400060008000R
espon
se
0123456789
YCH4 = 3.25%
(1-ε) = 0
(1-ε) = 0.05 (a) (b)
FT-IR
IR probe
91
Figure 6-11: Effect of (1-ε) and YCH4 on the variance of YCH4
6.2.4 Application in fluidized beds - chemical composition measurement in the
bubble phase
Gas tracer experiments were conducted in a fluidized bed and the fibre-optic probe was used with
the spectroscopic method to measure the molar fraction of a CH4 tracer inside the bubble and
emulsion phases. A bed of FCC particles was incipiently fluidized with nitrogen and bubbles
were produced at the tip of the fibre-optic probe by the injection of a nitrogen/methane mixture
(10.1% CH4) through a sparger. Figure 6.12 shows the history of solids volume fraction and
methane molar fraction measured by the fibre-optic probe during a typical experiment: the solids
volume fraction was 0.45 in the emulsion phase and decreased significantly to 0.05-0 when gas
bubbles of CH4/N2 were injected at intervals of approximately 11 seconds. The measured
methane volume fraction in the emulsion phase fluctuated significantly (5% to -5%) due to the
low intensity of the transmitted IR beam and the resulting low S/N. Figure 6.13(a) shows a
typical absorbance spectrum measured in the emulsion phase: the low intensity of reflected signal
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
3
4
56
78
910
0.000.02
0.040.06
0.08
σ
Y CH
4 (v
ol%
)
(1-ε)
92
resulted in a high absorbance throughout the spectrum. However, the time-average methane
molar fraction was measured accurately as 0.0% (100% nitrogen).
Figure 6-12: Simultaneous measurement of YCH4 and (1-ε) in the fluidized bed
Figure 6.13(b) shows an absorbance spectrum measured inside a bubble: a methane peak was
observed at 3017.63 cm-1. However, the injection of bubbles produced gas/solids movement,
which caused noise in most recorded absorbance spectra: 5 spectra out of the 27 measured in the
bubble phase were sufficiently clear to make a measurement of YCH4. Figure 6.13(c) shows a
typical absorbance spectrum where noise was observed and the methane peak was
indistinguishable. This noise could be eliminated in the wavelengths of interest by using a FT-IR
with a higher modulation frequency.
Time (s)
0 10 20 30 40 50
(1-ε
)
0.0
0.1
0.2
0.3
0.4
0.5
YC
H4 (
%)
-10
-5
0
5
10
15
20(1-ε)
YCH4
93
Figure 6-13: Absorbance spectra in fluidized bed
Wavenumber (cm-1
)
260028003000320034003600
A
-0.4
0.0
0.4
0.8
1.2
1.6
Wavenumber (cm-1
)
260028003000320034003600
A
-0.4
0.0
0.4
0.8
1.2
1.6
Wavenumber (cm-1
)
260028003000320034003600
A
-0.4
0.0
0.4
0.8
1.2
1.6(a)
(b)
(c)
CH4 peak
94
The average peak height measured in the bubble phase corresponded to a methane molar fraction
of 5% compared to the injected 10.1% CH4 in nitrogen. This discrepancy may be due to mixing
at the injector tip between the gas injected through the sparger and the fluidizing gas.
Furthermore, the small number of clear spectra that were measured in the bubble phase could also
explain the low value of YCH4 measured since time-average measurements were more accurate
than instantaneous measurements. During the present study, the measured absorbance spectra had
to be exported and analyzed manually, which greatly reduced the number of spectra that could be
considered to measure YCH4 in the bubble phase. However, these results clearly show that the
developed spectroscopic method with a fibre-optic probe can measure the gas composition in the
bubble phase.
6.2.5 Limitations of the technique and future work
This measurement technique is limited by the modulation frequency of the spectrometer: the
modulation frequency needs to be sufficiently high to avoid noise in the absorbance spectrum due
to movement in the gas/solid system. The FT-IR used in the present study was purchased in 2001
and new models are currently available with significantly higher modulation frequencies.
Another limitation of the technique is the IR beam intensity transmitted in the fibre-optic probe,
which needs to be as high as possible in order to maximize the signal-to-noise ratio. This is
critical if instantaneous measurements are required. On the other hand, time-averaged
measurements have been shown in the present study to be much more accurate with lower IR
beam intensity. The IR beam intensity can be increased by using a fibre-optic bundle, using a
more sensitive detector, using a more powerful IR source, optimizing the signal transmission at
the probe tip and at the FT-IR/fibre-optic probe interface (Harrick Fibremate™). More recent
models of FT-IR also propose higher IR signal intensity compared to the equipment used in the
present study.
95
The fibre-optic probe used in this study is limited to ambient temperature applications: fluoride
glass fibre-optics used can be exposed to temperatures below 150oC. However, fibre-optic probes
have been developed with plastic fibre-optics for temperatures up to 1000oC [41]. Therefore, it is
reasonable to assume that the measurement technique could be used to perform in-situ
measurements in high temperature multiphase systems with an adequately designed fibre-optic
probe.
6.2.6 Conclusions
A novel spectroscopic method was developed to measure quantitatively and simultaneously
solids volume fraction (1-ε) and gaseous species composition (Yi) in a gas/solid system. The
method was comprised of an FT-IR coupled to a fibre-optic probe that could perform real-time
and in-situ measurements of absorbance. The effect of solids volume fraction and gaseous
chemical composition on the absorbance spectra were additive and could be independently
calibrated. Experiments were conducted with methane/nitrogen and propane/nitrogen mixtures
and two types of particles: sand and FCC. Fuel mole fraction and solids volume fraction were
varied between 1.8 - 10.1 mol% and 0 - 0.45, respectively. The relative errors for the time-
averaged measurements of gaseous species mole fraction were below 6% and the error increased
significantly with decreasing beam intensity. The fibre-optic probe was also used in a fluidized
bed to measure the molar fraction of a gas tracer inside the emulsion and bubble phases during
gas tracer experiments.
The results strongly suggest that the accuracy of instantaneous fibre-optic probe measurements
could be significantly improved by increasing the IR beam intensity. This could be achieved by
using a fibre-optic bundle, using a more sensitive detector, using a more powerful IR source,
optimizing the signal transmission at the probe tip and at the FT-IR/fibre-optic probe interface
(Harrick Fibremate™). New FT-IR models also propose significantly higher IR signal intensity
compared to the equipment used in the present study. The effect of decreasing signal-to-noise
ratio could also be minimized by improving the curve smoothing technique to evaluate the
baseline absorbance.
96
More measurements need to be performed in gas/solid systems at ambient and high temperatures
to fully demonstrate the possibilities of this measurement method. High modulation frequencies
from newer models of FT-IR should help resolve the problem of noise in the absorbance spectra
caused by gas/solid movement. In theory, this method could also be used for liquid/solid,
gas/liquid, liquid/liquid and gas/liquid/solid systems. However, work has to be performed in
order to identify the limitations in these specific systems.
6.2.7 Acknowledgment
The authors are grateful to the Natural Sciences and Engineering Research Council of Canada
(NSERC) for their financial support. The authors would also like to thank IRphotonics for their
assistance in the fibre-optic probe construction.
6.2.8 Nomenclature
A Absorbance
dp Average particle size (µm)
f Modulation frequency (kHz)
t Distance between probe tip and mirror (mm)
Yi Molar fraction of specie i (vol%)
Greek symbols
(1−ε) Solids volume fraction
λ Wavelength (µm)
σ Variance
97
6.2.9 References
[1]Werther J, Molerus O. The local structure of gas fluidized beds – I. A statistically based
measuring system. International Journal of Multiphase Flow. 1973; 1 (1): 103-122.
[2]Yates JG, Simons SRJ. Experimental methods in fluidization research. International Journal
of Multiphase Flow. 1994; 20: 297-330.
[3]Hage B, Werther J. The guarded capacitance probe – A tool for the measurement of solids
flow patterns in a laboratory and industrial fluidized bed. Powder Technology. 1997; 93 (3): 235-
245.
[4]Begovich J, Watson JS. An electroconductivity technique for the measurement of axial
variation of hold-ups in three-phase fluidized beds. AIChE Journal. 1978; 24 (2): 351-354.
[5]Uribe-Salas A, Gomez CO, Finch JA. A conductivity technique for gas and solids holdup
determination in three-phase reactors. Chemical Engineering Science. 1994; 49 (1): 1-10.
[6]Liu M, Wang T, Yu W, Wang J. An electrical conductivity probe method for measuring the
local solid holdup in a slurry system. Chemical Engineering Journal. 2007; 132: 37-46.
[7]Huizenga P, Meijer R, Kuipers JAM, van Swaaij WPM. Fluoroptic solids holdup
measurements in slurry systems. AIChE Journal. 1998; 44 (4): 982-988.
[8]Cutolo A, Rendina I, Arena U, Marzocchella A, Massimilla L. Optoelectronic technique for
the characterization of high concentration gas-solid suspension. Applied Optics 1990, 29 (9),
pp. 1317-1322.
[9]Ishida M, Tanaka H. Optical probe to detect both bubbles and suspended particles in a three-
phase fluidized bed. Journal of Chemical Engineering of Japan. 1982; 15 (5): 389-391.
[10]Hu T, Yu B, Wang Y. Holdups and models of three phase fluidized beds. In: Ostergaard K,
Sorensen A. Fluidization V. New York: Engineering Foundation, 1986: 353.
[11]Esmaeili B, Chaouki J, Dubois C. An evaluation of the solid hold-up distribution in a
fluidized bed of nanoparticles using radioactive densitometry and fibre optics. Canadian Journal
of Chemical Engineering. 2008; 86: 543-552.
[12]Gandhi B, Prakash A, Bergougnou MA. Hydrodynamic behaviour of slurry column at high
solids concentrations. Powder Technology. 1999; 103 (2): 80-94.
98
[13]Chaouki J, Larachi F, Duduković MP. Noninvasive tomographic and velocimetric
monitoring of multiphase flows. Industrial & Engineering Chemistry Research. 1997; 36 (11):
4476-4503.
[14]Pugsley T, Tanfara H, Malcus S, Cui H, Chaouki J, Winters C. Verification of fluidized bed
electrical capacitance tomography measurements with a fibre optic probe. Chemical Engineering
Science. 2003; 58 (17): 3923-3934.
[15]Marashdeh Q, Fan LS, Du B, Warsito W. Electrical capacitance tomography – A perspective.
Industrial & Engineering Chemistry Research. 2008; 47 (10): 3708-3719.
[16]Razzak SA, Barghi S, Zhu JX, Mi Y. Phase holdup measurements in a gas-liquid-solid
circulating fluidized bed (GLSCFB) riser using electrical resistance tomography and optical fibre
probe. Chemical Engineering Journal. 2009; 147: 210-218.
[17]Uchida S, Okamura S, Katsumata T. Measurement of longitudinal distribution of solids
holdup in a three-phase fluidized bed by ultrasonic technique. Canadian Journal of Chemical
Engineering. 1989; 67: 166-169.
[18]Luypaert J, Massart DL, Vander Heyden Y. Near-infrared spectroscopy applications in
pharmaceutical analysis. Talanta. 2007; 72 (3): 865-883.
[19]Swarbrick B. Process analytical technology: a strategy for keeping manufacturing viable in
Australia. Vibrational Spectroscopy. 2007; 44: 171-178.
[20]Armaroli T, Bécue T, Gautier S. Diffuse reflection infrared spectroscopy (drifts) : application
to the in situ analysis of catalysts. Oil & Gas Science and Technology. 2004 ; 59 (2): 215-237.
[21]Kondo JN, Yang S, Zhu Q, Inagaki S, Domen K. In situ infrared study of n-heptane
isomerization over Pt/H-beta zeolites. Journal of Catalysis. 2007; 248 (1): 53-59.
[22]Corveleyn S, Vandenbossche GMR, Remon JP. Near-infrared (NIR) monitoring of H2O2
vapor concentration during vapour hydrogen peroxide (VHP) sterilisation. Pharmaceutical
Research. 1997; 14 (3): 294-298.
[23]Rantanen J, Räsänen E, Tenhunen J, Känsäkoski M, Mannermaa J-P, Yliruusi J. In-line
moisture measurement during granulation with a four-wavelength near infrared sensor: an
99
evaluation of particle size and binder effects. European Journal of Pharmaceutics and
Biopharmaceutics. 2000a; 50 (2): 271-276.
[24]Rantanen J, Antikainen O, Mannermaa J-P, Yliruusi J. Use of the near-infrared reflectance
method for measurement of moisture content during granulation. Pharmaceutical Development
and Technology. 2000; 5 (2): 209-217.
[25]Cogdill RP, Anderson CA, Delgado M, Chisholm R, Bolton R, Herkert T, Afnan AM,
Drennen III JK. Process analytical technology case study part I: feasibility studies for quantitative
near-infrared method development. AAPS PharmSciTech. 2005; 6 (2): E273-E283.
[26]Triadaphillou S, Martin E, Montague G, Norden A, Jeffkins P, Stimpson S. Fermentation
process tracking through enhanced spectral calibration modeling. Biotechnology and
Bioengineering. 2007; 97 (3): 554-567.
[27]Andersson M, Josefson M, Langkilde FW, Wahlund K-G. Monitoring of a film coating
process for tablets using near infrared reflectance spectrometry. Journal of Pharmaceutical and
Biomedical Analysis. 1999; 20: 27-37.
[28]Kirsch JD, Drennen JK. Determination of film-coated tablet parameters by near-infrared
spectroscopy. Journal of Pharmaceutical and Biomedical Analysis. 1995; 13: 1273-1281.
[29]Kirsch JD, Drennen JK. Near-infrared spectroscopic monitoring of the film coating process.
Pharmaceutical Research. 1996; 13 (2): 234-237.
[30]Abrahamsson C, Johansson J, Andersson-Engels S, Svanberg S, Folestad S. Time-resolved
NIR spectroscopy for quantitative analysis of intact pharmaceutical tablets. Analytical Chemistry.
2005; 77 (4): 1055-1059.
[31]Hailey PA, Doherty P, Tapsell P, Oliver T, Aldridge PK. Automated system for the on-line
monitoring or powder blending processes using near-infrared spectroscopy part I. system
development and control. Journal of Pharmaceutical and Biomedical Analysis. 1996; 14
(5): 551-559.
[32]Sonja Sekulic S, Wakeman J, Doherty P, Hailey PA. Automated system for the on-line
monitoring of powder blending processes using near-infrared spectroscopy: Part II. Qualitative
approaches to blend evaluation Journal of Pharmaceutical and Biomedical Analysis. 1998;
17: 1285-1309.
100
[33]Sonja Sekulic S, Ward HW, Brannegan DR, Stanley ED, Evans CL, Sciavolino ST, Hailey
PA, Aldridge PK. On-line monitoring of powder blend homogeneity by near-infrared
spectroscopy. Analytical Chemistry. 1996; 68: 509-513.
[34]Frake P, Greenhalgh D, Grierson SM, Hempenstall JM, Rudd DR. Process control and end-
point determination of a fluid bed granulation by application of near infra-red spectroscopy.
International Journal of Pharmaceutics. 1997; 151 (1): 75-80.
[35]Bellamy LJ, Nordon A, Littlejohn D. Real-time monitoring of powder mixing in a convective
blender using non-invasive reflectance NIR spectrometry. The Analyst. 2008; 133: 58-64.
[36]Cui H, Mostoufi N, Chaouki J. Comparison of measurement techniques of local particle
concentration for gas-solid fluidization. In: Kwauk M, Li J, Yang W-C. Fluidization X. New
York: Engineering Foundation, 2001: 779-786.
[37]De Paepe ATG, Dyke JM, Hendra PJ, Langkilde FW. Rotating samples in FT-RAMAN
spectrometers. Spectrochimica Acta Part A. 1997; 53: 2261-2266.
[38]Berntsson O, Danielsson L-G, Folestad S. Characterization of diffuse reflectance fiber probe
sampling on moving solids using a Fourier transform near-infrared spectrometer. Analytical
Chimica Acta. 2001; 431: 125-131.
[39]Andersson M, Svensson O, Folestad S, Josefson M, Wahlund K-G. NIR spectroscopy on
moving solids using a scanning grating spectrometer – impact on multivariate process analysis.
Chemometrics and Intelligent Laboratory Systems. 2005; 75: 1-11.
[40]Skoog DA, Holler FJ, Crouch SR. Principles of Instrumental Analysis (6th edition). Brooks
Cole. 2006.
[41]Cui H, Sauriol P, Chaouki J. High temperature fluidized bed reactor: measurements,
hydrodynamics and simulation. Chemical Engineering Science. 2003; 58: 1071-1077.
101
CHAPITRE 7 DISCUSSION GÉNÉRALE
Afin d’éviter la combustion dans la zone de désengagement, il est important que la conversion
des réactifs soit la plus élevée possible dans la région de lit fluidisé de solides afin d’éviter la
formation d’une mixture inflammable en aval. La température du lit fluidisé est un des
paramètres clé qui détermine le taux de conversion dans cette région. Le premier article de cette
thèse (chapitre 4) a mesuré expérimentalement les températures critiques de transitions T1
(température du lit fluidisé en-dessous de laquelle le front de combustion est situé dans la zone de
désengagement) et T2 (température au-dessus de laquelle la combustion se produit à quelques
centimètres du point d’injection).
Les résultats démontrent qu’une température de lit de 1073 K (T2) est suffisante pour que la
combustion des n-alkanes C2-C4 se produise entièrement dans le lit et à quelques centimètres du
point d’injection. Si la température est diminuée en dessous de 923 K (T1), le front de combustion
se trouvera au-dessus de la surface du lit fluidisé. Pour le méthane, ces températures sont
significativement plus élevées avec T1 = 1023 K et T2 > 1123 K.
Ces températures critiques pourraient varier si les conditions d’opérations changent. Par exemple,
une vitesse de fluidisation plus élevée diminuerait le temps de résidence du gaz. D’un autre côté,
le taux de transfert de masse entre les phases bulle et émulsion augmenterait avec un changement
de régime de fluidisation (bullage à turbulent, par exemple) et pourrait favoriser une combustion
rapide. En conséquence, les températures critiques mesurées dans cette étude devraient, a priori,
constituer une bonne indication pour un large éventail de conditions de fluidisation. Toutefois, il
serait intéressant d’effectuer des mesures similaires à des vitesses superficielles plus élevées.
La cinétique de réaction globale a également été mesurée pour le méthane, l’éthane, le propane et
le n-butane. Cette mesure a été effectuée an assumant que l’hydrodynamique de la phase gazeuse
était bien caractérisée par deux réacteurs pistons en série: un réacteur piston pour la région de
l’injecteur et un autre pour la région développée du lit fluidisé. La validité de cette hypothèque
semble confirmée par les observations de Lorences et al. (2006) et aussi par la bonne
102
concordance du modèle de combustion développé dans cette étude avec les données
expérimentales. La cinétique de combustion mesurée pour l’éthane, le propane et le n-butane
pouvait être exprimée par une seule expression de type Arrhenius. Toutefois, il faut noter que la
cinétique est peut-être légèrement différente, mais à l’intérieur de la marge d’erreur des mesures.
Cette similitude entre les cinétiques de réactions est possiblement due à des agents de réactions
radicalaires en chaîne similaires entre les n-alcanes C2-C4. Pour le méthane, la cinétique de
combustion mesurée était significativement plus lente. Il faut noter que le transfert de masse
n’était pas limitant lors de ces expériences. Dans un tel cas, l’énergie d’activation aurait été
proche de zéro puisque que les coefficients de diffusion augmentent à peu près linéairement en
fonction d’une hausse de température.
À proximité de l’injecteur, le taux de transfert de matière est significativement plus élevé
comparativement au lit fluidisé. Les résultats expérimentaux ont démontré que les conditions
dans la région de l’injecteur et dans le lit fluidisé peuvent produire une conversion élevée à
proximité de l’injecteur ou un délai d’autoallumage dans la région du lit fluidisé. La conversion
mesurée près de l’injecteur augmentait avec une hausse de la température du lit et diminuait avec
un accroissement de la vitesse du jet à l’extrémité de l’injecteur. L’effet inhibiteur de la vitesse
d’injection sur les réactions de combustion a également été observé lors d’études sur l’auto-
allumage dans des écoulements à contre-courant (Fotache et al. 1999, Jomaas et al. 2005,
Humer et al. 2002). Afin d’inclure les effets promoteur et inhibitifs de la région de l’injecteur, un
parallèle a été effectué avec le concept de temps d’induction: la longueur du réacteur piston
caractérisant la zone de l’injecteur a été corrélée à la température du lit fluidisé et de la vitesse
d’injection avec une expression de type Arrhenius. Une telle expression peut prédire de manière
précise le temps d’induction d’une mixture inflammable dans un système homogène. Le modèle
de réaction démontre une bonne concordance avec les résultats expérimentaux.
Lorsque des contraintes opératoires limitent la conversion des réactifs dans la région du lit
fluidisé, il est important d’évaluer le risque d’auto-allumage et de réactions dans la zone de
désengagement. Le deuxième article de cette thèse (chapitre 5) constitue une étude expérimentale
sur la combustion du propane dans la zone de désengagement. Lors de ces expériences, la vitesse
103
du gaz de fluidisation était maintenue près de la vitesse minimale de fluidisation afin de
minimiser l’emportement de solides en aval du lit de solides. Des profils axiaux de fractions
volumétriques de propane ont été mesurés à la surface d’un lit fluidisé et dans la zone de
désengagement. La température du lit fluidisé était maintenue faible (818 K - 923 K) afin que la
combustion de propane s’effectue essentiellement dans la zone de désengagement. Les mesures
ont été effectuées après que l’opération du réacteur ait atteint un régime permanent. Il est
intéressant de noter que l’auto-allumage de la mixture inflammable de propane/air se produisait
toujours à la surface du lit fluidisé et jamais en aval. Ceci peut être expliqué par le fait que la
température dans la région de désengagement atteignait sa valeur maximale à la surface du lit
fluidisé lorsque l’injection du propane débutait. Sinon, cela peut mettre en évidence un couplage
entre la zone de réaction du lit fluidisé et celle de la zone de désengagement – la réaction se
produisant dans le lit fluidisé se poursuivait dans la zone de désengagement.
Les résultats expérimentaux ont été comparés à un modèle de réaction utilisant un mécanisme de
microcinétique pour un système homogène en phase gazeuse. La microcinétique est essentielle
afin de reproduire la production de radicaux libres qui précède l’autoallumage. Six mécanismes
distincts de microcinétique ont été utilisés et comparés. Les résultats suggèrent que les agents de
réactions radicalaires en chaîne produits dans le lit fluidisé peuvent être entraînés dans la zone de
désengagement où ils accélèrent la combustion. Dans le présent, les modèles suggèrent que du
peroxyde d’hydrogène est l’agent de réaction radicalaire principale, ce qui est également affirmer
dans la littérature scientifique pour les températures utilisées.
Lors des expériences sur la combustion dans le lit fluidisé (chapitre 4) et dans la zone de
désengagement (chapitre 5), la caractérisation des régions de faibles fractions de solides a été très
ardue sinon impossible. Afin de caractériser la zone de désengagement, une mesure simultanée de
flux de particules et de composition chimique a été effectuée à l’aide d’une sonde à prélèvement
non-isocinétique. Or, la mesure était moyennée dans le temps puisque les mesures instantanées
étaient impossibles avec cette méthode. De plus, il n’est pas simple de convertir une mesure de
flux de solides en fraction de solides afin de considérer l’effet des particules sur la cinétique. Ces
expériences auraient grandement bénéficiés d’une méthode de mesure simultanée de la fraction
104
de solides et de la composition chimique de la phase gazeuse. Dans la région du lit fluidisé, la
caractérisation des réactions se produisant la région à proximité de l’injecteur et dans la phase
bulle aurait nécessitée une technique similaire.
Dans le chapitre 6, une méthode de mesure spectroscopique permettant la mesure simultanée de
la fraction de solides et de la composition chimique de la phase gazeuse a été développée. En
mode transmission, les résultats ont démontré que l’absorbance produite par des solides en
suspension et celle produite par une composition chimique sont additives. Ainsi, cette
caractéristique rend possible la mesure simultanée de ces deux paramètres en mesurant des
spectres d’absorbance. Or, ceci ne serait probablement pas valable en mode diffusion diffuse
puisque le volume de mesure dépendrait de la fraction de solides.
La méthode spectroscopique a été utilisée avec une sonde à fibres optiques infrarouges afin
d’effectuer des mesures en ligne (fréquence de mesure de 4.5 Hz) et in situ dans un écoulement
gaz/solide et un lit fluidisé. Des mesures d’une grande précision ont été effectuées avec une
erreur relative généralement de moins de 5% sur la composition chimique. Également, la
méthode a été utilisé avec succès pour mesurer la fraction volumétrique d’un traceur dans la
phase bulle d’une lit fluidisé gaz/solide.
Les résultats démontrent que la méthode est limitée par deux paramètres:
• L’intensité du rayon infrarouge transmis dans l’échantillon et récolté par le détecteur du
spectromètre.
• La vitesse de mouvement des solides et la fréquence de modulation du spectromètre.
L’intensité du signal mesuré par le spectromètre et le ratio signal sur bruit (S/N) décroissait en
fonction de la diminution du signal infrarouge. Un signal faible augmentait l’erreur sur les
mesures instantanées. L’erreur relative sur la mesure pouvait toutefois être significativement
105
diminuée en-dessous de 5% si une moyenne temporelle des mesures instantanées était effectuée.
L’intensité du rayon infrarouge pourrait être accrue en:
• Augmentant l’intensité de la source infrarouge
• Optimisant le transfert du rayon infrarouge à l’interface spectromètre/sonde à fibres-
optiques
• Optimisant le transfert du rayon infrarouge l’extrémité de la sonde
• Augmentant la sensibilité du détecteur
De nouveaux modèles de FT-IR sont caractérisés par des rayons infrarouge plus puissants. De
plus, un plus grand nombre de fibres optiques ou des fibres plus grosses pourraient être utilisé.
Donc, il y a des avenues possibles afin d’augmenter l’intensité du signal infrarouge.
Le mouvement des particules solides produisait du bruit dans le spectre d’absorbance mesuré.
Afin d’effectuer des mesures dans des systèmes multiphasiques, les résultats du chapitre 6
démontrent que la fréquence de modulation doit être suffisamment élevée afin de déplacer le
bruit aux basses fréquences et à l’extérieur des longueurs d’ondes d’intérêts. Lors des présentes
expériences, la fréquence de modulation était suffisamment élevée afin d’éliminer le bruit.
Toutefois, pour des applications où la vitesse des particules solides serait plus élevée, de
nouveaux modèles de FT-IR offrent présentement des fréquences de modulation supérieures aux
80 kHz utilisés dans cette étude.
La sonde infrarouge développée ne peut résister à des températures au-delà d’environ 150oC.
Toutefois, des sondes à fibres-optiques ont été conçues afin de résister à des températures
atteignant 1000oC (Cui et al. 2003).
106
CONCLUSION
L’objectif principal de cette thèse était de développer des outils de modélisation et de
caractérisation des phénomènes de réactions homogènes en phases gazeuses dans les réacteurs à
lit fluidisé.
Dans la première partie de ce travail, la combustion en lit fluidisé du méthane, éthane, propane et
n-butane avec de l’air a été étudiée en mode d’injection séparée. Un lit fluidisé de particules
inertes de sables a été opéré dans le régime à bulle et à des températures intermédiaires :
923 K ≤ TB ≤ 1123 K. Pour l’éthane, le propane et le n-butane, la combustion avait lieu
principalement dans la région de désengagement lorsque la température du lit fluidisé était
inférieure à T1 = 923 K. D’un autre côté, la combustion se produisant entièrement à l’intérieur
d’une distance de 0.2 m de l’injecteur lorsque la température du lit était supérieure à T2 = 1073 K.
Pour le méthane, les valeurs de T1 et T2 étaient significativement plus élevées avec des mesures
respectives de 1023 K et 1123 K. La combustion dans le lit fluidisé a été modélisée avec
précision à l’aide d’une cinétique globale de première ordre et une modèle hydrodynamique
composé de deux réacteurs pistons en séries : un qui caractérisait la région près de l’injecteur et
un autre pour la région de lit fluidisé en aval. La cinétique globale de réaction a été mesurée pour
chacun des hydrocarbures. Pour les n-alcanes C2 à C4, la cinétique globale était caractérisé par
une seule expression de type Arrhenius, alors que la cinétique du méthane était significativement
plus lente. Les propriétés de la région de l’injecteur causaient soit une conversion importante
dans cette région, soit un délai d’auto-allumage dans le lit fluidisé. La conversion des réactifs
près de l’injecteur augmentait avec une hausse de la température du lit fluidisé et diminuait avec
une hausse de la vitesse d’injection du gaz à l’extrémité de l’injecteur. Afin de prendre ce
phénomène en compte dans le modèle de réaction, une analogie a été effectuée avec le concept de
temps d’induction. Ainsi, la longueur du réacteur piston caractérisant la région de l’injection a été
corrélée à la température du lit fluidisé et la vitesse d’injection avec une expression de type
Arrhenius.
107
Dans la seconde partie de cette thèse, la combustion du propane dans la zone de désengagement
d’un lit fluidisé de sable a été étudiée. La température du lit fluidisé a été maintenue faible entre
818 K et 923 K afin que la combustion se produise essentiellement en aval du lit fluidisé. De
plus, une vitesse superficielle de gaz deux fois supérieure à la vitesse minimale de fluidisation a
été utilisée pour minimiser l’emportement des solides. Durant les expériences, la zone de
désengagement a été caractérisée par des mesures simultanées de flux de solides, de composition
chimique de la phase gazeuse, de température et de pression. L’auto-allumage de la mixture
propane/air a été observé à seulement 0.06 m de la surface du lit pour des températures de lit
fluidisé supérieures à 833 K. Six mécanismes de microcinétique ont été utilisés afin de reproduire
le profil axial de conversion du propane: tous les modèles ont sous-estimé le taux de réaction
dans la zone de désengagement. Or, lorsque le modèle considérait le H2O2 produit lors de la
combustion dans le lit fluidisé, la vitesse de réaction calculée dans la zone de désengagement
étaient significativement plus élevée et menait à une meilleure concordance entre le modèle et les
résultats expérimentaux.
La troisième partie de ce travail relate le développement d’une nouvelle méthode spectroscopique
pour la mesure quantitative et simultané de la fraction de solides (1-ε) et de la composition
chimique dans la phase gazeuse (Yi) dans un système gaz/solide. La méthode a été utilisée avec
un FT-IR connecté à une sonde à fibres-optiques à laquelle des mesures en temps réel et in situ
ont été effectuées. L’effet de (1-ε) et Yi sur le spectre d’absorbance mesuré étaient additifs et
pouvaient être calibrés de manière indépendante. Des mesures ont été effectuées avec des
mixtures alkane/azote et deux types de particules: du sable et du catalyseur FCC. La fraction
volumétrique de l’hydrocarbure et (1-ε) ont été variées entre des valeurs respectives de
1.8 - 10.1 mol% et 0 - 0.45. Les erreurs relatives sur la mesure de Yi étaient toujours inférieures à
6% and l’erreur augmentait avec une baisse de l’intensité du rayon infrarouge. Une preuve de
concept a été effectuée dans un lit fluidisé: la sonde à fibres-optiques a été utilisée pour mesurer
la fraction volumétrique d’un traceur gazeux dans la phase bulle et émulsion.
Les résultats ont également démontré que la méthode est limitée par deux paramètres:
108
• L’intensité du rayon infrarouge transmis dans l’échantillon et récolté par le détecteur du
spectromètre.
• La vitesse de mouvement des solides et la fréquence de modulation du spectromètre.
Un signal infrarouge faible augmente l’erreur sur les mesures instantanées. L’erreur relative sur
la mesure peut toutefois être significativement diminuée en-dessous de 5% si une moyenne
temporelle est effectuée. L’intensité du rayon infrarouge pourrait être accrue en:
• Augmentant l’intensité de la source IR
• Optimisant le transfert du rayon IR à l’interface spectromètre/sonde à fibres-optiques
• Optimisant le transfert du rayon IR l’extrémité de la sonde
• Augmentant la sensibilité du détecteur
Le mouvement des particules solides produit du bruit dans le spectre d’absorbance mesuré de
sorte que la fréquence de modulation doit être suffisamment élevée afin de déplacer le bruit vers
les basses fréquences à l’extérieur des longueurs d’ondes d’intérêts.
109
BIBLIOGRAPHIE
Abrahamsson C., Johansson J., Andersson-Engels S., Svanberg S., Folestad S. (2005). Time-
resolved NIR spectroscopy for quantitative analysis of intact pharmaceutical tablets. Analytical
Chemistry, 77(4), 1055-1059.
Andersson M., Josefson M., Langkilde F.W., Wahlund K.G.. (1999) Monitoring of a film coating
process for tablets using near infrared reflectance spectrometry. Journal of Pharmaceutical and
Biomedical Analysis, 20, 27-37.
Andersson M., Svensson O., Folestad S., Josefson M., Wahlund K.G.. (2005). NIR spectroscopy
on moving solids using a scanning grating spectrometer – impact on multivariate process
analysis. Chemometrics and Intelligent Laboratory Systems, 75, 1-11.
Armaroli T., Bécue T., Gautier S. (2004). Diffuse reflection infrared spectroscopy (drifts) :
application to the in situ analysis of catalysts. Oil & Gas Science and Technology, 59(2), 215-
237.
Baron J., Bulewicz E.M., Zukowski W., Kandefer S., Pilawska M.. (2002). Combustion of
hydrocarbon fuels in a bubbling fluidized bed. Combustion and Flame, 128, 410-421.
Begovich J., Watson J.S.(1978). An electroconductivity technique for the measurement of axial
variation of hold-ups in three-phase fluidized beds. AIChE Journal, 24 (2): 351-354.
Bellamy L.J., Nordon A., Littlejohn D. (2008). Real-time monitoring of powder mixing in a
convective blender using non-invasive reflectance NIR spectrometry. The Analyst, 133, 58-64.
Benjelloun F., Liegeois R., Vanderschuren J. (1995). Penetration length of horizontal gas jets into
atmospheric fluidized beds. In J.F. Large, C. Laguerie, Fluidization VIII: Proceedings of the
Eighth Engineering Foundation Conference on Fluidization (American Institute of Chemical
Engineers, 239–246). New York: Engineering Foundation.
Berntsson O., Danielsson L.G., Folestad S. (2001). Characterization of diffuse reflectance fiber
probe sampling on moving solids using a Fourier transform near-infrared spectrometer.
Analytical Chimica Acta, 431, 125-131.
110
Botella P., López Nieto J.M., Solsona B. (2002). Selective oxidation of propene to acrolein on
Mo-Te mixed oxides catalysts prepared from ammonium telluromolybdates. Journal of
Molecular Catalysis A:Chemical, 184, 335-347.
Brokaw R.S., Jackson J.L. (1955). Effect of temperature, pressure, and composition on ignition
delays for propane flames. The 5th International Symposium on Combustion, (563-569). New
York: Reihold.
Buda F., Glaude P.A., Battin-Leclerc F., Porter R., Hughes K.J., Griffiths J.F. (2006). Use of
detailed kinetic mechanisms for the prediction of autoignitions, Journal of Loss Prevention in the
Process Industries, 19, 227-232.
Burcat A., Scheller K., Lifshitz A. (1971). Shock-tube investigation of comparative ignition delay
times for C1-C5 alkanes. Combustion and Flame, 16, 29-33.
Burcat A., Lifshitz A. (1971). Shock-tube investigation of ignition in propane-oxygen-argon
mixtures. Proceedings of the Combustion Institute, 745-755.
Burcat A., Scheller K., Lifshitz A. (1971). Shock-tube investigation of comparative ignition delay
times for C1-C5 alkanes. Combustion and Flame, 16, 29-33.
Cadman P., Thomas G.O., Butler P.. (2000). The auto-ignition of propane at intermediate
temperature and high pressures. Physical Chemistry Chemical Physics, 2, 5411-5419.
Cathonnet M., Boettner J.C., James H. (1981). Experimental study and numerical modeling of
high temperature oxidation of propane and n-butane. Proceedings of the 18th Symposium of the
Combustion Institute, (18, 903-913). Pittsburgh: The Combustion Institute.
Chang C.J., Thompson A.L., Winship R.D. (1959). Ignition delay of propane in air between 725-
880oC under isothermal conditions. Proceedings of the 7h Symposium of the Combustion Institute,
(431-435). Pittsburgh: The Combustion Institute.
Chaouki J., Klvana D., Guy C.. (1999). Selective and complete catalytic oxidation of natural gas
in turbulent fluidized beds. Korean Journal of Chemical Engineering, 16(4), 494-500.
Chaouki J., Larachi F., Duduković M.P. (1997). Noninvasive tomographic and velocimetric
monitoring of multiphase flows. Industrial & Engineering Chemistry Research, 36(11), 4476-
4503.
111
Chen L.H., Wen C.Y. (1982). Model of solid gas reaction phenomena in the fluidized bed
freeboard. AIChE Journal, 28(6), 1019-1027.
Cogdill R.P., Anderson C.A., Delgado M., Chisholm R., Bolton R., Herkert T., Afnan A.M.,
Drennen III J.K. (2005). Process analytical technology case study part I: feasibility studies for
quantitative near-infrared method development. AAPS PharmSciTech, 6(2), E273-E283.
Corveleyn S., Vandenbossche G.M.R., Remon J.P. (1997). Near-infrared (NIR) monitoring of
H2O2 vapor concentration during vapour hydrogen peroxide (VHP) sterilisation. Pharmaceutical
Research, 14(3), 294-298.
Cui H., Mostoufi N., Chaouki J. (2001). Comparison of measurement techniques of local particle
concentration for gas-solid fluidization. In Kwauk M., Li J., Yang W.C. Fluidization X. (779-
786). New York: Engineering Foundation.
Cui H., Sauriol P., Chaouki J. (2003). High temperature fluidized bed reactor: measurements,
hydrodynamics and simulation. Chemical Engineering Science, 58, 1071-1077.
Cutolo A., Rendina I., Arena U., Marzocchella A., Massimilla L. (1990). Optoelectronic
technique for the characterization of high concentration gas-solid suspension. Applied Optics, 29
(9), 1317-1322.
Dagaut P., Cathonnet M., Boettner J.C., Gaillard F. (1987). Kinetic oxidation of propane
oxidation. Combustion Science and Technology, 56, 23-63.
De Lasa H.I., Grace J.R. (1979). The influence of the freeboard region in a fluidized bed catalytic
cracking regenerator. American Institute of Chemical Engeneers Journal, 25(6), 984-991.
Dennis J.S., Hayhurst A.N., Mackley I.G. (1982). The ignition and combustion of propane/air
mixtures in a fluidised bed. The 19th International Symposium on Combustion, (1205-1212).
Pittsburgh: The Combustion Institute.
De Paepe A.T.G., Dyke J.M., Hendra P.J., Langkilde F.W. (1997). Rotating samples in FT-
RAMAN spectrometers. Spectrochimica Acta Part A, 53, 2261-2266.
Dounit S., Hemati M., Andreux R. (2008). Modelling and experimental validation of a fluidized-
bed reactor freeboard region: application to natural gas combustion. Chemical Engineering
Journal, 140, 457-465.
112
Esmaeili B., Chaouki J., Dubois C. (2008). An evaluation of the solid hold-up distribution in a
fluidized bed of nanoparticles using radioactive densitometry and fibre optics. Canadian Journal
of Chemical Engineering, 86, 543-552.
Foka M., Chaouki J., Guy C., Klvana D. (1994). Natural gas combustion in a catalytic turbulent
fluidized bed. Chemical Engineering Science, 49(24A), 4269-4276.
Fotache C.G., Wang H., Law C.K. (1999). Ignition of ethane, propane, and butane in counterflow
jets of cold fuel versus hot air under variable pressures. Combustion and Flame, 117, 777-794.
Frake P., Greenhalgh D., Grierson S.M., Hempenstall J.M., Rudd D.R. (1997). Process control
and end-point determination of a fluid bed granulation by application of near infra-red
spectroscopy. International Journal of Pharmaceutics, 151(1), 75-80.
Freeman G., Lefebvre A.H. (1984). Spontaneous ignition characteristics of gaseous hydrocarbon-
air mixtures. Combustion and Flame, 58, 153-162.
Friedman J., Li H. (2005). Natural gas combustion in a fluidized bed heat-treating furnace.
Combustion Science and Technology, 177, 2211-2241.
Gandhi B., Prakash A., Bergougnou M.A. (1999). Hydrodynamic behaviour of slurry column at
high solids concentrations. Powder Technology, 103(2), 80-94.
Hage B., Werther J. (1997). The guarded capacitance probe – A tool for the measurement of
solids flow patterns in a laboratory and industrial fluidized bed. Powder Technology, 93(3), 235-
245.
Hailey P.A., Doherty P., Tapsell P., Oliver T., Aldridge P.K. (1996). Automated system for the
on-line monitoring or powder blending processes using near-infrared spectroscopy part I. system
development and control. Journal of Pharmaceutical and Biomedical Analysis, 14(5), 551-559.
Hartman M., Trnk O., Pohořelý M., Svoboda K.(2010). High-temperature reaction in the
freeboard region above a bubbling fluidized bed. Industrial & Engineering Chemistry Research,
49(6), 2672–2680.
Hesketh R.P., Davidson J.F. (1991). Combustion of methane and propane in an incipiently
fluidized bed. Combustion and Flame, 85, 449-467.
113
Hong R., Li H., Li H., Wang Y. (1997). Studies on the inclined jet penetration length in a gas–
solid fluidized bed, Powder Technology, 205–212.
Horning D.C., Davidson D.F., Hanson R.K. (2002). Study of the high-temperature autoignition of
n-alkane/O2/Arixtures. Journal of Propulsion and Power, 18(2), 363-371.
Hu T., Yu B., Wang Y. (1986). Holdups and models of three phase fluidized beds. In Ostergaard
K., Sorensen A. Fluidization V. (353). New York: Engineering Foundation.
Huizenga P., Meijer R., Kuipers J.A.M., Van Swaaij W.P.M. (1998). Fluoroptic solids holdup
measurements in slurry systems. AIChE Journal, 44 (4), 982-988.
Humer S., Seiser R., Seshadri K. (2002). Non-premixed and premixed extinction and autoignition
of C2H4, C2H6, C3H6 and C3H8. Proceedings of the Combustion Institute, 29, 1597-1604.
Hutchenson K.W., La Marca C., Patience G.S., Laviolette J.P., Bockrath R.E. (2010). Parametric
study of butane oxidation in a circulating fluidized bed reactor. Catalysis A: General, 376(1-2),
91-103.
Ishida M., Tanaka H. (1982). Optical probe to detect both bubbles and suspended particles in a
three-phase fluidized bed. Journal of Chemical Engineering of Japan, 15(5), 389-391.
Jeng R.S., Altwicker E.R., Morgan III M.M., Wilk R.D. (2001). Propane combustion in a spouted
bed combustor II: a modeling study in the bed region. Combustion Science and Technology, 170,
119-149.
Jomaas G., Zheng X.L., Zhu D.L., Law C.K. (2005). Experimental determination of counterflow
ignition temperatures and laminar flame speeds of C2-C3 hydrocarbons at atmospheric and
elevated pressures. Proceedings of the Combustion Institute, 30, 193-200.
Kin K., Soo Shin K. (2001). Shock tube and modeling study of the ignition of propane. Bulletin
of the Korean Chemical Society, 22(3), 303-307.
Kirsch J.D., Drennen J.K. (1995). Determination of film-coated tablet parameters by near-
infrared spectroscopy. Journal of Pharmaceutical and Biomedical Analysis, 13, 1273-1281.
Kirsch J.D., Drennen J.K. (1996). Near-infrared spectroscopic monitoring of the film coating
process. Pharmaceutical Research, 13(2), 234-237.
114
Kondo J.N., Yang S., Zhu Q., Inagaki S., Domen K. (2007). In situ infrared study of n-heptane
isomerization over Pt/H-beta zeolites. Journal of Catalysis, 248(1), 53-59.
Kunii D., Levenspiel O. (1990). Entrainment of solids from fluidized beds I.Hold-up of solids in
the freeboard II.Operation of fast fluidized bed. Powder Technology, 61, 193-206.
Lamoureux N., Paillard C.E., Vaslier V. (2002). Low hydrocarbon mixtures ignition delay times
investigation behind reflected shock waves. Shock Waves, 11, 309-322.
Laviolette J.P., Sotudeh-Gharebagh R., Mabrouk R., Patience G.S., Chaouki J. (2010) Fluidized
bed combustion of methane and other hydrocarbons. In Agarwal A. Handbook of Combustion
(vol.5), Weinheim: Wiley.
Liu M., Wang T., Yu W., Wang J. (2007). An electrical conductivity probe method for measuring
the local solid holdup in a slurry system. Chemical Engineering Journal, 132, 37-46.
Lorences M., Laviolette J.P. (2006). Fluid bad gas RTD: effect of fines and internals. Powder
Technology, 168(1), 1-9.
Lorences M.J., Patience G.S., Cenni R., Diez F., Coca J.. (2006a). VPO transient lattice oxygen
contribution. Catalysis Today, 112, 45-48.
Luypaert J., Massart D.L., Vander Heyden Y. (2007). Near-infrared spectroscopy applications in
pharmaceutical analysis. Talanta, 72(3), 865-883.
Mabrouk R., Chaouki J., Laviolette J.P., Patience G.S. (2008). On natural gas combustion in a
fluidized bed. Proceedings of the 9th Conference on Circulating Fluidized Beds, Hamburg,
Germany.
Malleswara Rao T.V., Deo G. (2007). Ethane and Propane Oxidation over Supported V2O5/TiO2
Catalysts: Analysis of Kinetic Parameters. Industrial & Engineering Chemistry Research, 46, 70-
79.
Marashdeh Q., Fan L.S., Du B., Warsito W. (2008). Electrical capacitance tomography – A
perspective. Industrial & Engineering Chemistry Research, 47(10), 3708-3719.
Marinov N.M., Pitz W.J, Westbrook C.K., Hori M., Matsunaga N. (1998). An experimental and
kinetic calculation of the promotion effect of hydrocarbons on the NO-NO2 conversion in a flow
reactor. Proceedings of the Combustion Institute, 27, 389-396.
115
Merry J.M.D. (1971). Penetration of a horizontal gas jet into a fluidised bed. Transactions of the
Institution of Chemical Engineers, 49(4), 189–195.
Okhotskii V.B. (2001). Depth of penetration of a gas jet into a granular medium. Theoretical
Foundations of Chemical Engineering, 35(2), 199-200.
Penyazkov O.G, Ragotner K.A., Dean A.J., Varatharajan B. (2005). Autoignition of propane-air
mixtures behind reflected shock waves. Proceedings of the Combustion Institute, 30, 1941-1947.
Pre P., Hemati M., Marchand B. (1998). Study on natural gas combustion in fluidized beds:
modelling and experimental validation. Chemical Engineering Science, 53(16), 2871-2883.
Pugsley T., Tanfara H., Malcus S., Cui H., Chaouki J., Winters C. (2003). Verification of
fluidized bed electrical capacitance tomography measurements with a fibre optic probe. Chemical
Engineering Science, 58(17), 3923-3934.
Qin Z., Lissianski V.V., Yang H., Gardiner Jr. W.C., Davis S.G., Wang H. (2000). Combustion
chemistry of propane: a case study of detailed reaction mechanism optimization. Proceedings of
the Combustion Institute, 28, 1663–1669.
Rantanen J. Antikainen O. Mannermaa J.P., Yliruusi J. (2000). Use of the near-infrared
reflectance method for measurement of moisture content during granulation. Pharmaceutical
Development and Technology, 5(2), 209-217.
Rantanen J., Räsänen E., Tenhunen J., Känsäkoski M., Mannermaa J.P., Yliruusi J.(2000a). In-
line moisture measurement during granulation with a four-wavelength near infrared sensor: an
evaluation of particle size and binder effects. European Journal of Pharmaceutics and
Biopharmaceutics, 50(2), 271-276.
Ranzi E., Faravelli T., Gaffuri P., Pennati G.C., Sogaro A. (1994). A wide range modeling study
of propane and n-butane oxidation. Combustion Science and Technology, 100, 299-330.
Razzak S.A., Barghi S., Zhu J.X., Mi Y. (2009) Phase holdup measurements in a gas-liquid-solid
circulating fluidized bed (GLSCFB) riser using electrical resistance tomography and optical fibre
probe. Chemical Engineering Journal, 147, 210-218.
Reinhardt, B., Cordonnier, A., Florent, P. (1999). Use of an isokinetic sampling propane. Results
in a cyclone. Powder Technology, 101, 81-90.
116
Rhodes, M..J., Laussmann, P. (1992). A simple non-isokinetic sampling probe for dense
suspensions. Powder Technology, 70, 141-151.
Ross D.P., Yan H.M., Zhang D.K. (2001). An experimental study of propane combustion in a
fluidized-bed gasifier. Combustion and Flame, 124, 156-164.
Ross D.P., Yan H.M., Zhang D.K. (2004). Modelling of a laboratory-scale bubbling fluidised-bed
gasifier with feeds of both char and propane. Fuel, 83, 1979-1990.
Saxena P., Peters N., Williams F.A. (2007). An analytical approximation for high-temperature
autoignition times of higher alkanes. Combustion and Flame, 149, 79-90.
Shakhova N.A. (1968). Discharge of turbulent jets into a fluidized bed, Journal of Engineering
Physics and Thermophysics, 14 (1), 32–36.
Shen Z., Briens C.L., Kwauk M., Bergougnou M.A. (1990). Study of a downward gas jet in a
two-dimensional fluidized bed. Canadian Journal of Chemical Engineering, 68, 534-540.
Singh B., Rigby G.R., Callcott T.G. (1975). Combustion of gaseous fuels in a fluidised bed.
Fluidised Combustion Conference. London: Paper C5.
Skoog D.A., Holler F.J., Crouch S.R.(2006) Principles of Instrumental Analysis (6th edition).
Belmont: Brooks Cole.
Smith G.P., Golden D.M., Frenklach M., Moriarty N.W., Eiteneer B., Goldenberg M., Bowman
C.T., Hanson R.K., Song S., Gardiner Jr. W.C., Lissianski V.V., Qin
Z. http://www.me.berkeley.edu/gri_mech/
Sonja Sekulic S., Wakeman J., Doherty P., Hailey P.A. (1998) Automated system for the on-line
monitoring of powder blending processes using near-infrared spectroscopy: Part II. Qualitative
approaches to blend evaluation Journal of Pharmaceutical and Biomedical Analysis,
17, 1285-1309.
Sonja Sekulic S., Ward H.W., Brannegan D.R., Stanley E.D., Evans C.L., Sciavolino S.T., Hailey
P.A., Aldridge P.K. (1996). On-line monitoring of powder blend homogeneity by near-infrared
spectroscopy. Analytical Chemistry, 68, 509-513.
Sotudeh-Gharebaagh R. (1998). Combustion of natural gas in a turbulent fluidized bed reactor.
PhD Thesis, École Polytechnique, Montréal, Québec, Canada.
117
Sotudeh-Gharebagh R., Chaouki J. (2003). The heterogeneous and homogeneous combustion of
methane over inert particles. Canadian Journal Chemical Engineering , 81, 1182-1191.
Sotudeh-Gharebaagh R., Chaouki J. (2007). Investigation of highly exothermic reactions in a
turbulent fluidized bed reactor. Energy & Fuels, 21, 2230-2237.
Sotudeh-Gharebagh R., Mostoufi N. (2003). Simulation of a catalytic turbulent fluidized bed
reactor using the sequential modular approach. Fuel Processing Technology, 85, 189-200.
Srinivasan R.A., Sriramulu S., Kulasekaran S., Agarwal P.K. (1998). Mathematical modeling of
fluidized bed combustion-2: combustion of gases. Fuel, 77, 1033-1049.
Stubington J.F., Davidson J.F. (1981). Gas-phase combustion in fluidized beds. AIChE Journal,
27 (1), 59-65.
Sung C.J., Li B., Law C.K., Wang H. (1998). Structure and sooting limits in counterflow
methane/air and propane/air diffusion flames from 1 to 5 atmospheres, The 27th International
Symposium of Combustion, (1523-1530). Pittsburgh: The Combustion Institute.
Swarbrick B. (2007). Process analytical technology: a strategy for keeping manufacturing viable
in Australia. Vibrational Spectroscopy, 44, 171-178.
Triadaphillou S., Martin E., Montague G., Norden A., Jeffkins P., Stimpson S. (2007)
Fermentation process tracking through enhanced spectral calibration modeling. Biotechnology
and Bioengineering, 97(3), 554-567.
Uchida S., Okamura S., Katsumata T. (1989) Measurement of longitudinal distribution of solids
holdup in a three-phase fluidized bed by ultrasonic technique. Canadian Journal of Chemical
Engineering, 67, 166-169.
Uribe-Salas A., Gomez C.O., Finch J.A. (1994). A conductivity technique for gas and solids
holdup determination in three-phase reactors. Chemical Engineering Science, 49(1), 1-10.
Van der Vaart D.R. (1988). Freeboard ignition of premixed hydrocarbon gas in a fluidized bed.
Combustion and Flame, 71, 35-39.
Van der Vaart D.R. (1988a). The chemistry of premixed hydrocarbon/air combustion in a
fluidized bed. Fuel, 67, 1003-1007.
118
Van der Meer, E.H., Thorpe, R.B. & Davidson, J.F. (2000). Flow patterns in the square cross-
section riser of a circulating fluidized bed and the effect of riser exit design. Chemical
Engineering Science, 55, 4079-4099.
Walsh P.M., Chaung T.Z., Dutta A., Beér J.M., Srofim A.F. (1982). Particle entrainment and
nitric oxide reduction in the freeboard of a fluidized coal combustor. ACS Fuels Volumes, 27(1),
243-261.
Wen C.Y., Chen L.H. (1982). Fluidized bed freeboard phenomena: entrainment and elutriation.
AIChE Journal, 28(1), 117-128.
Werther J., Molerus O. (1973). The local structure of gas fluidized beds – I. A statistically based
measuring system. International Journal of Multiphase Flow, 1(1), 103-122.
Westbrook, C.K. (2000). Chemical kinetics of hydrocarbon ignition in practical combustion
systems. Proceedings of the Combustion Institute, 28, 1563-1577.
Wilk R.D., Cernansky N.P., Cohen R.S. (1986). The oxidation of propane at low and transition
temperatures. Combustion Science and Technology, 49, 41-78.
Yates J.G., Simons S.R.J. (1994) Experimental methods in fluidization research. International
Journal of Multiphase Flow, 20, 297-330.
Zenz F.A. (1968). Bubble formation and grid design, IChemE Symposium Series, 30, 136–139.
Zhukov V.P., Sechenov V.A. Starikovskii. (2005). Autoignition of a lean propane-air mixture at
high pressures. Kinet Catal, 46(3), 319-327.
Zukowski W. (2003). A simple model for explosive combustion of premixed natural gas with air
in a bubbling fluidized of inert sand. Combustion and Flame, 134, 399-409.
Chemical-Kinetic Mechanisms for Combustion Applications, San Diego Mechanism web page,
Mechanical and Aerospace Engineering (Combustion Research), University of California at San
Diego. (http://maeweb.ucsd.edu/~combustion/cermech/)
119
ANNEXE 1 – ARTICLE PRESENTÉE À LA CONFÉRENCE
CHEMCON 2006 (INDE)
Propane-oxygen induction time measurement in the freeboard of a
fluidized bed – Freeboard characterization
Jean-Philippe Laviolette1a
, Julia Duchastel-Légaréa, Gregory S. Patience
a, Concetta La Marca
b
and Jamal Chaoukia
aDepartment of Chemical Engineering, Ecole Polytechnique, Montreal, Quebec,Canada
bDuPont Engineering Research and Technology, Wilmington, DE, USA
ABSTRACT
In the present study, the gas and solids hydrodynamics were characterized inside the freeboard and
disengagement sections of a pilot scale fluidized bed using gas RTD and a fibre-optics probe for two
superficial velocities: Ug = 0.2 and 0.6 m/s. Experiments were performed at ambient temperature and
vanadyl pyrophosphate catalyst (Geldart group A) was used as the bed material. Results showed that the
hydrodynamic regime inside the freeboard was close to plug flow for both velocities. However, the
diverging area leading to the disengagement region produced large scale fluctuations in the Gas RTD at
low velocity, which disappeared at higher velocity. Furthermore, the fibre-optic probe showed that the
solids concentration was approximately constant along the radius of the reactor and decreased with
increasing height as well as decreasing superficial velocity.
INTRODUCTION
The direct oxidative conversion of lower alkanes to oxygenated products is undergoing intense research
due to their lower costs compared to olefins. Propane to acrylic acid is one such process and several
kinetic studies [1] have indicated that high oxygen (15-70%) and low propane (2-3%) concentrations
favour the selective reaction. Furthermore, while propane conversion increases with temperature, reaction
1 Present address: Department of Chemical Engineering, Ecole Polytechnique, C.P. 6079, succ. Centre-
Ville, Montreal (Quebec), Canada, H3C 3A7 ([email protected])
120
temperatures above approximately 450oC were reported to result in a loss of selectivity. Under these
conditions, the risk of homogeneous gas-phase auto-ignition represents a safety concern. In fluidized bed
reactors, explosive mixtures may be avoided by feeding oxygen and hydrocarbons to the bed separately.
However, under partial conversion conditions, the catalyst concentration in the freeboard region is
insufficient to quench free radicals and thus gas phase auto-ignition becomes an important safety hazard.
The risk of auto-ignition in the freeboard, cyclones and downstream piping must be quantified for design
purposes.
Induction time – the time in which a hydrocarbon-oxygen mixture auto-ignites – is one essential design
parameter. Propane-oxygen induction time measurements can be found throughout the literature [2].
However, agreement between these studies is generally lacking and the discrepancy most likely is due to
the strong relationship between the induction time and the experimental conditions as well as vessel
geometry. Consequently, there remains a need to develop generalized relationships between propane-
oxygen induction time, process conditions, solids concentration and reactor geometry at different scales.
Previous studies have shown that induction time varies with temperature (T), pressure (P), mixture
composition, solids concentration (CS) and vessel characteristics but extrapolating these tendencies to a
fluidized bed introduces uncertainties. Induction time studies have been performed in several types of
vessels (static vessels, flow reactors, shock tubes and high compression machines), but never in fluidized
beds.
Induction times also are likely dependant upon the gas hydrodynamics. Previous studies performed in flow
reactors were at high turbulent flow regimes such that perfect radial mixing was assumed. Chang et al. [3]
observed no significant variations as the Reynolds number was varied from 2000 to 14000. However, if
induction time is to be measured in a fluidized bed under realistic operating conditions, the flow regime
may be closer to laminar and the gas-phase hydrodynamics may have two essential effects. First, it may
affect the induction time directly since the mass transfer will determine the mixture to wall heat transfer,
which may prevent thermal run-away. Secondly, it may give rise to radial gradients that will render the
contribution of the different physical parameters (T, P, CS, etc) difficult to assess. In the end, this
potentially can lead to significant uncertainties in the induction time correlation. Hence, the gas-phase
hydrodynamics must be characterized before proceeding to the measurement of induction time.
Solids concentration is another factor which has been shown to affect induction time. In a fluidized bed
operating at velocities above the minimum bubbling velocity, bubbles form at the distributor and rise
through the bed. As the bubbles reach the bed surface and collapse, particles are dispersed into the
freeboard and disengagement regions. The effect of solids has never been quantified in induction time
121
measurement studies nor included in correlations. Furthermore, radial gradients in solids concentrations
will lead to uncertainties in the system characterization. It is therefore essential to characterize the solids
hydrodynamics inside the freeboard and disengagement regions.
In order to relate the position of auto-ignition (detectable with a thermocouple) to the induction time, the
average mixture residence time must be measured at the different radial and axial locations inside the
reactor. This can easily be done with gas RTD studies. In prior studies, gas RTD experiments have been
mostly performed inside the bed of solids [4]. In comparison, the gas hydrodynamics freeboard and
disengagement regions should be simpler due to the absence of bubbles, but it is complicated by the
presence of a diverging area. In fact, the effect of a disengagement region on gas RTD not been
previously reported.
In the present study, experiments were performed in order to characterize the gas and solids
hydrodynamics inside the freeboard (I.D. = 20 cm) and disengagement (I.D. = 60 cm) regions of a
fluidized bed reactor at ambient temperature and with equilibrated VPO (Geldart A) as the bed material.
Gas RTD studies were performed to determine the flow regime of the gas above the bed. RTD was also
used to map the average gas residence time at different positions (axial and radial) inside the two regions
of interest. Lastly, a fibre-optic probe was used to quantify the variation of solids concentration with radial
and axial positions inside the freeboard.
EXPERIMENTAL APPARATUS
The experiments were conducted in a fluidized bed as shown in Figure 1. The bed and freeboard regions
had an inner diameter of 20 cm, while the disengagement section had an inner diameter of 60 cm. The
experimental procedure consisted of first fluidizing the solids with air at ambient temperature for all
experiments. In the RTD studies, propane was used as the tracer and was injected as a step function into
the fluidizing air stream upstream of the windbox. Preliminary tests showed that the tracer mixed perfectly
before reaching the windbox. The RTD curves for different radial and axial positions were then obtained
through in-line sampling and analysis with an FT-IR unit with a recording speed of 2 Hz. The tracer was
injected via the opening of a solenoid four-way valve which was triggered by the FT-IR unit. This resulted
in a variability of the measured average gas residence time that was below 0.8 s in the freeboard region.
Furthermore, the volumetric sampling rate was set at 5 mL/s which represented less than 1% of the total
flow inside the reactor and minimized the perturbations on the system.
122
Figure 1: Experimental apparatus
To quantify the solids concentration into the freeboard region, a fibre-optics probe was used. Essentially,
the probe consisted of 4 emitting and 4 receiving fibre-optics with a 1mm thick window at the tip such that
the fibre-optic output was linearly proportional to the solids concentrations. A PV-4A particle velocity
analyzer (Institute or Process Engineering, Chinese Academy of Sciences) was used as a light source
and to (record time series of instantaneous local voidage at various axial and radial positions. To ensure
the validity and the repeatability of the sampled signals, five series of 33 000 data points were collected at
a sampling rate of 2 kHz for every position.
RESULTS AND DISCUSSION
The freeboard region begins at the bed surface and it is continuously moving due to bubble bursting. The
bubbles also entrain solids into the freeboard resulting in a vertical gradient of solids concentration.
Consequently, the bed surface gets diffused and does not constitute a well defined line.
Solids concentration
The bed surface can be more accurately defined by measuring pressure records. Figure 2 shows the
measured pressure profile inside the reactor for two superficial velocities as well as the calculated values
given by a modified Kato and Wen model [5]. Since the pressure drop is proportional to the void fraction,
60 cm
20 cm
97 cm
20 cm
48 cm
C3H8
Air
Disengagement
region
Bed and freeboard
regions
FT-IR
Sampling
probe
Air
Z r
123
we can assume that the freeboard region begins where the pressure reaches a plateau and the solids
fraction is low. Hence, at Ug = 0.6 m/s, it is observed from the experimental data and the model that the
bed surface is located somewhere between 35 and 45 cm. Also, at Ug = 0.16 m/s, the bed surface is
located slightly lower at around 35 cm.
Figure 2: Axial pressure profile in the fluidized bed reactor
A fibre-optic probe was used to quantify the variation of solids fraction (1-ε) with axial and radial positions
and the results are shown in Figures 3 and 4. Hence, Figure 3 shows that the solids concentration was
approximately invariant with radial position. Furthermore, we note from Figure 4 that the radially averaged
solids concentration remained constant near the bed surface and then decreased with increasing axial
position and decreasing superficial velocity. Near the bed, the bubble bursting phenomenon governs the
solids entrainment, which results in a region of high solids fraction. However, as Z increases, the amount
of solids ejected by the bubbles decreases and the gas superficial velocity gradually becomes a more
important factor in entrainment. The experimental data was compared to the results reported by
Morooka et al. [6] for FCC catalyst (dp = 66 µm) at Ug = 0.15 and 0.5 m/s. The results of Morooka et al. [6]
show that the exponential decrease of (1-ε) with Z was valid only at a certain height (approximately 75 cm)
above the bed surface. Our results agree fairly well with their data.
Gas-phase hydrodynamics in the freeboard and disengagement regions
To characterize the gas-phase hydrodynamics we used gas RTD with step function gas injection.
Therefore, the first derivative of the experimental data (F(t) curve) was taken to obtain the E(t) curve.
Hence, Figure 5 shows the E(t) curve inside the freeboard for Ug = 0.2 m/s (a) and 0.6 m/s (b) at three
Z (cm)
0 20 40 60 80 100
Ga
ge
pre
ssu
re (
kP
a)
0
1
2
3
4
5
6
7
8
Ug = 0.16 m/s
Ug = 0.6 m/s
Modified Kato & Wen model
124
radial positions. It is seen that the curves for each of the three different radial positions are
indistinguishable and have a nearly Gaussian shape. Therefore, the gas was well mixed radially. This was
observed at all positions inside the freeboard.
Figure 3: Solids volume fraction w.r.t. axial and radial positions
(Filled symbols [Ug = 0.6 m/s] and empty symbols [Ug = 0.16 m/s])
Figure 4: Average solids fraction w.r.t. to axial position
r/R
0.0 0.2 0.4 0.6 0.8 1.0
(1-ε
)
10 -3
10 -2
10 -1
100
Z = 35 cm
Z = 50 cm
Z = 75 cm
Z (cm)
20 40 60 80 100 120 140 160 180 200
(1-ε
)
10-4
10-3
10-2
10-1
100
Z vs Ug = 0.16 m/s
Z vs Ug = 0.6 m/s
Morooka et al. [6]
125
Figure 5: E(t) curve in the freeboard at Z = 96 cm and for: (a) Ug = 0.6 m/s and (b) Ug = 0.2 m/s taken at
r/R = 0, 0.5, 1.
However, inside the disengagement region and at Ug = 0.2 m/s, the E(t) curve deformed as shown in
Figure 6 which was obtained at the centreline at two axial positions. As the superficial velocity was
increased to 0.6 m/s, the E(t) curve retained its Gaussian shape as observed from Figure 7. At low
velocity and inside the freeboard, the flow regime is close to turbulent flow with a Reynolds number (Re)
of 2750, which is near the commonly used threshold of 3000. However, in the disengagement region, due
to an increase in diameter (3×) and a decrease in Ug (÷9 by assuming rapid expansion of the gas), the Re
is divided by a factor of three and the flow becomes laminar. Therefore, there are most likely large scale
mixing patterns that form inside the disengagement section and deforms the tracer front. However, at
Ug = 0.6 m/s, the flow is highly turbulent inside the reactor with Re = 8250 and it approaches plug flow.
Table 1 lists the first (tmean) and second (σ2) moments of the E(t) curve for different axial positions inside
the freeboard and disengagement regions. Note that the reported values have been averaged over the
entire section of the reactor. In essence, Table 1 shows that the variance does not change significantly
with height inside the freeboard, which further supports the conclusion that the hydrodynamics in that
region is close to plug flow. This is also true for Ug = 0.6 m/s inside the disengagement region.
0
0.02
0.04
0.06
0.08
0.1
0 20 40 60
Time (s)
E(t)
0
0.02
0.04
0.06
0 30 60 90Time (s)
E(t)
(a) (b)
126
Figure 6: E(t) curve for Ug = 0.2 m/s and for: (a) Z = 122 cm and (b) Z = 152 cm at the centreline (r/R = 0)
Table 1: Properties of the E(t) curves and n-CSTRs model parameters
Z (cm) Ug (cm/s) σ2 (s
2) tmean (s) (tmean-t)max (s)
45 20 85.8 42.1 0.2
96 20 90.5 43.4 0.8
122 20 88.9 51.9 1.6
152 20 103.3 70.4 3.8
96 60 22 31.9 0.4
122 60 20.3 31.2 0.8
152 60 20.7 34.6 0.6
In order to relate the position of ignition to the induction time, the average residence time of the gas must
be measured. Furthermore, the fifth column reports the maximum absolute variation. It is shown that
inside the freeboard and disengagement (at Ug = 0.6 m/s only) regions, the variation reaches a maximum
of 0.8 seconds. On the other hand, the variation increased significantly at Ug = 0.2 m/s inside the
0
0.02
0.04
0.06
0 20 40 60 80 100
Time (s)
E(t)
0
0.02
0.04
0.06
0 30 60 90 120
Time (s)
E(t)
(a) (b)
127
disengagement region due to the large scale mixing patterns. This larger variation will lead to higher
uncertainties in the induction time measurements.
Figure 7: E(t) curve for Ug = 0.6 m/s and for: (a) Z = 122 cm and (b) Z = 152 cm taken at r/R = 0, 0.5, 1.
CONCLUSIONS
In the present study, the solids and gas hydrodynamics were characterized. First, it was shown with fibre-
optic probe measurements that the solids concentration remained approximately constant with radial
position and with axial position close to the bed surface. As the axial position was increased, the solids
fraction decreased with both increasing height and decreasing superficial velocity. Furthermore, gas RTD
experiments showed that the hydrodynamic regime inside the freeboard was close to plug flow at Ug = 0.2
and 0.6 m/s velocities. However, the diverging area leading to the disengagement region produced
perturbations on the E(t) curve at low velocity which were likely due to large scale eddies in the flow since
the Reynolds number was inside the laminar range. At higher gas velocity (Ug = 0.6 m/s), the flow
approached plug flow again inside the disengagement region.
REFERENCES
[1] Landi, G., Lisi, L. and Volta, J.-C. “Oxidation of propane to acrylic acid over vanadyl pyrophosphate:
modifications of the structural and acid properties during the precursor activation and their relationship
with catalytic performance”, Journal of Molecular Catalysis A: Chemical, 222, 2004, P. 175-181.
[2] Herzler, J., Jerig, L. and Roth, P. “Shock-tube study of the ignition of propane at intermediate
temperatures and high pressures”, Combustion Science and Technology, 176, 2004, P. 1627-1637.
0
0.02
0.04
0.06
0.08
0.1
0 20 40 60
Time (s)
E(t)
0
0.02
0.04
0.06
0.08
0.1
0.12
0 20 40 60
Time (s)E
(t)
(a) (b)
128
[3] Chang, C.J., Thompson, A.L. and Winship, R.D. “Ignition delay of propane in air between 725-800oC
under isothermal conditions”, 7th International Symposium on Combustion (1958, London). Proceedings.
The Combustion Institute, editor. Pittsburgh. Combustion Institute, 1958. P. 431-435.
[4] Lorences, M.J., Laviolette, J.-P., Patience, G.S., Alonso, M. and Diez, F.V. “Fluid bed gas RTD: Effect
of fines and internals”, Powder Technology, 168, 2006, P. 1-9.
[5] Kato, K., Wen,C. Y., "Bubble assemblage model for fluidised bed catalytic reactor" , Chemical
Engineering Science, 24, 1969, P. 1351-1369.
[6] Morooka, S., Kawazuishi, K. and Kato, Y. “Holdup and flow pattern of solid particles in freeboard of
gas-solid fluidized bed with fine particles”, Powder Technology, 26, 1980, P. 75-82.
129
ANNEXE 2 – ARTICLE PRESENTÉE À LA CONFÉRENCE IFSA 2008
Gas-phase combustion in the freeboard of a fluidized bed
Jean-Philippe Laviolette, Gregory S. Patience and Jamal Chaouki
Ecole Polytechnique, Montréal, Canada
Keywords: fluidized bed, propane, combustion, freeboard, non-isokinetic sampling, solids flux
ABSTRACT
Propane was injected inside a high temperature fluidized bed through a sparger and the freeboard
region was characterized using pressure, solids flux, species volume fraction and temperature
measurements during freeboard combustion. Solids flux and species volume fractions were
measured simultaneously using a non-isokinetic probe. A decrease in propane and oxygen as well
as an increase in temperature was observed in the freeboard. The solids flux measurements
showed a net upward flow at the centre of the reactor and a net downward flow near the wall.
Furthermore, the solids entrainment was found to decrease exponentially with height from the
bed surface. The measured entrainment at the bed surface was significantly higher than predicted
by the correlation of Wen et al. [1]: 17.6 kg/m2s compared to 8.18×10-4 kg/m2s. Pressure and
solids flux measurements were also used to locate the fluidized bed surface. Solids flux
measurements were more precise than pressure to determine bed height. The location of the bed
surface was taken as the inflection point of the entrainment curve. Finally, the temperature profile
corresponding to the measured solids flux and species volume fractions was determined using a
series of thermocouples positioned along the length of the reactor.
INTRODUCTION
Many industrial catalytic processes involve simultaneous heterogeneous and homogeneous
reaction pathways. Homogeneous reactions generally result in yield lost of the desired
compound, but they may also represent a significant safety hazard in the case of fast and highly
130
exothermic reactions such as those involving oxygen and hydrocarbons. Chemical processes with
oxygen and hydrocarbons are widespread in the industry. Furthermore, new processes are
currently under development in the context of oil price fluctuations and global warming: selective
oxidation of alkanes, combustion of cheaper feedstock for heating and catalyst regeneration,
biomass gasification, etc. Fluidized bed reactors are currently being developed for a variety of
these processes. In order to maximize productivity, they may operate within the explosion limits
while feeding the oxidant and hydrocarbon separately into the bed – through spargers. This is
possible due to the ability of the solids phase to suppress the homogeneous reactions
(combustion) and at the same time promote selective heterogeneous reactions.
Due to operational constraints, these processes may operate under partial conversion conditions.
For example, selectivity generally declines at high conversion in selective oxidation processes
due to an increase in the rate of series reactions [2, 3, 4]. The implications for fluid bed is that
downstream of the solids – in the freeboard, cyclones and associated piping – the effluent gas
phase composition is potentially explosive – high hydrocarbon and oxygen concentrations,
elevated temperature with an insufficient solids volume fraction to quench non-selective
homogeneous reactions [5, 6]. Minimizing the risk of gas-phase combustion-deflagration in these
regions remains an important design issue.
The threshold conditions leading to autoignition (self-initiated homogeneous combustion) must
be known to minimize the safety hazards not only during standard operation but also during start-
up as well as emergency shut down conditions. Several explosion criteria (explosion limits,
autoignition temperature and induction times [6]) measured in homogeneous gas-phase systems
as well as gas-phase kinetic models are available in the scientific literature [7]. However,
extrapolating these tendencies to a fluidized bed reactor introduces uncertainties since the
entrained solid particles affect the gas-phase reactions downstream of the fluidized bed.
Furthermore, fluidized beds are non-homogeneous systems - solids entrainment decreases
exponentially from the bed surface to the transport disengagement height [1]. Also, an axial
temperature gradient may exist in the freeboard region as well as species concentration gradients
depending on the hydrodynamics.
131
To accurately predict gas-phase combustion in the freeboard of a fluidized bed for a wide range
of operational conditions, a freeboard combustion model that combines the gas/solids
hydrodynamics and the reaction kinetics must be developed. This model needs to integrate
through the axial gradient of solids fraction and temperature that characterize the freeboard as
well as accounting for the effects of solids on the reaction kinetics. Models of the reaction in the
freeboard of fluidized beds have been previously proposed by Chen et al. [8] Dounit et al. [9] and
Dounit et al. [10]. These models used entrainment models that include a correlation for the
entrainment at the bed surface such as the model of Wen et al. [1]. However, the solids
entrainment at the bed surface was extrapolated from experimental data measured several
centimetres downstream of the bed surface (sometimes more than 50 centimetres). Therefore, the
existing entrainment models may yield significant errors near the bed surface and better accuracy
would be obtained by directly measuring solids fraction.
The species volume fractions and solids fraction measured in the freeboard are two dependent
parameters since the reaction kinetics varies with solids fraction. The ability of sand particles to
inhibit homogeneous reactions and to delay autoignition has been shown by Hesketh et al. [6].
Solids entrainment has been previously measured using non-isokinetic probes [11, 12, 13]. The
measured solids flux using such probes has been found to be constant for a range of sampling
rates such that the probe does not necessarily have to be at the isokinetic condition. Also, solids
fraction has been measured using fibre-optic probes [14, 15] and capacity probes [15]. With these
techniques, solids fraction/entrainment is measured independently of species volume fractions.
In the present study, propane was injected inside a high temperature fluidized bed through a
downward facing sparger and the freeboard region was characterized using pressure, solids flux,
species volume fraction and temperature measurements during freeboard combustion. The
fluidized bed temperature, superficial velocity and propane/air ratio were adjusted to obtain
propane combustion in the freeboard. Solids flux and species volume fractions were measured
simultaneously using a non-isokinetic probe. The temperature profile in the freeboard was
monitored using eight thermocouples.
132
EXPERIMENTAL SETUP & CONDITIONS
The experiments were conducted in the fluidized bed reactor of Figure 1. The bed and freeboard
regions had an inner diameter of 20 cm, while the disengagement section had an inner diameter
of 60 cm. For all experiments, sand particles (Geldart group B – ρs = 2650 kg/m3, dp = 250 µm)
were used as the bed material and air was the fluidizing media. The experimental procedure first
consisted of adjusting the superficial gas velocity to 24 cm/s (at 650oC). The fluidized bed was
then heated to 680oC with the natural gas burner upstream of the windbox and by injecting heavy
vacuum gas oil (HVGO) through the sparger. When the bed temperature reached 680oC, the
burner and HVGO injection were stopped and propane was injected through the sparger for 30
minutes to burn HVGO residues.
Figure 1: Fluidized bed reactor
The propane was then turned off and the fluidized bed cooled down until the bed temperature
reached 645oC. Propane was then injected through the downward sparger to obtain a mixed-cup
3 4
Non-isokinetic sampling probe
0.48 m
0.2 m
0.97 m
0.38 m
1 1st cyclone
2 2nd
cyclone 3 Natural gas burner 4 Windbox 5 Fluidized bed
0.6 m
0.2 m
Air + Natural gas
C3H8
r
5
O2 analyser
CO/CO2 analyser
z
1
2
Gas
chromatograph
133
volume fraction of 2 vol% in air. A non-isokinetic probe was used to sample gas and solids
simultaneous from 5 axial and 6 radial positions. The sampling system is shown in Figure 2 and
consisted of a 4.1 mm I.D. probe bent at a 90o at its tip. The sampled gas/solid stream first went
through a 250 mL container that lowered the sampled gas velocity and caused the separation of
the solid particles from the gas. The gas was analyzed by a Siemens Ultramat CO/CO2 analyzer
and a Varian CP-4900 micro-GC that measured the following species: O2, H2, CH4, C2H4, C2H6,
C3H6, C3H8, CO, CO2 and N2. Solids flux was calculated from the mass of the sampled solids and
the sampling time:
tA
mG DSU ., = (1)
The cross-sectional area at the probe’s tip was 0.66 cm2. The upward (GSU) and downward (GSD)
solid fluxes were measured by orienting the probe tip downward and upward, respectively. The
sampling flow rate was controlled using a rotameter connected to a vacuum line. During the
sampling time, the bed temperature increased to 650ºC.
Gas was also sampled in the disengagement region (z = 147 cm) and analyzed with an oxygen
analyzer (Siemens Oxymat 6.1). The results from the GC, CO/CO2 analyzer and oxygen analyzer
were compared and used in a molar balance to confirm the validity of the measured species
volume fractions. The temperature along the reactor was monitored using 8 thermocouples
located along the length of the reactor. The sparger consisted of a 4.1 mm I.D. tube pointing
downward at the bed centreline. The sparger tip was located at a height of 9.8 cm above the
distributor.
134
Figure 2: Schematic of sampling system
RESULTS AND DISCUSSION
The reaction rate of propane combustion in the freeboard of a fluid bed depends on species
volume fractions, temperature and solids fraction. They are dependent parameters and should
ideally be measured simultaneously.
Simultaneous sampling of gas and solids
A non-isokinetic sampling probe was used to measure the species volume fractions and the solids
flux simultaneously. From the superficial gas velocity (Ug = 24 cm/s) and the probe cross-
sectional area, the sampling rate required to obtain a gas velocity of 24 cm/s at the probe tip was
310 mL/min. However, a preliminary set of experiments was conducted to determine the effect of
gas sampling rate on the solids flux. The sampling probe was positioned near the bed at
z = 29 cm and the upward solids flux (GSU) was measured at three radial positions: r = 0 (centre),
4 and 8 cm. The gas sampling rate was varied from 232 mL/min to 580 mL/min. These
experiments were conducted at a bed temperature of 650ºC and with propane freeboard
combustion. Upward solids flux as a function of gas sampling rate is shown in Figure 3(a): the
measured solids flux was almost independent of sampling rate in the range of 270 - 470 mL/min.
Therefore, it was not necessary to have sampling that was precisely iso-kinetic and a sampling
rate of 425 mL/min was used for all experiments.
Solids recovery
GC
&
CO/CO2
analyser
Non-isokinetic probe To
vacuum
135
Figure 3: The isokinetic condition
Experiments were also conducted to determine the effect of sampling time on the measured solids
flux. Figure 3(b) shows the measured upward solids flux at the centreline as a function of
sampling time. The solids flux decreased by 14% (from 35.7 kg/m2s to 31.4 kg/m2s) as the
sampling time was increased from 31 to 60 seconds. After 60 seconds, the measured solids flux
remained more constant with sampling time: it decreased by 7.0 % (from 31.4 kg/m2s to
29.2 kg/m2s) as the sampling time was increased to 141 seconds. This decrease in measured
solids flux was most probably due to the fact that the mass of sampled solids represented a
significant fraction of the total bed: 275 g was sampled after 141 seconds, which represented
3.3 % of the total mass of the bed (8.3 kg). These results suggest that a minimum sampling time
of 60 seconds was required to reach a permanent condition.
The gas sampled with the non-isokinetic sampling probes was sent to the CO/CO2 analyzer and
the GC for analysis. Figure 4 shows the measured CO2 species fractions at the centreline
(r = 0 cm), z = 87 cm and with Q = 425 mL/min as a function of sampling time. A sampling time
of only 20 seconds was required for a representative measurement of the species volume fractions
inside the reactor.
Sampling time (s)
20 40 60 80 100 120 140 160
GS
U (kg/m
2.s
)
0
10
20
30
40
Q (mL/min)
200 300 400 500 600
GS
U (kg/m
2.s
)
0
10
20
30
40
0
4
8
z = 29 cm
r = 0 cm
r (cm)
Q = 425 mL/min(a) (b)
136
Figure 4: Sampling time impact
Solids flux in the freeboard
Upward (GSU) and downward (GSD) solid fluxes were measured at 5 axial positions and 6 radial
positions. The radial profiles of solids flux are shown in Figure 5(a) for three axial positions: 29,
35 and 45 cm above distributor. The upward (empty symbols) and downward (filled symbols)
solids fluxes reached a maximum at reactor centreline and decreased closer to the wall. The
measurements were averaged over the fluid bed cross-section. At z = 29 cm, the upward and
downward solids fluxes were 17.6 kg/m2s and 19.0 kg/m2s, respectively (difference of
1.4 kg/m2s). At z = 35 cm, the upward solids flux was 4.7 kg/m2s and the downward solids flux
was 3.0 kg/m2s (difference of 1.7 kg/m2s). At z = 45 cm, the upward and downward solids flux
were both zero.
The radial profiles of net solids flux (GS) are shown in Figure 5(b) for three axial positions: there
was a net upward of solids at the centre of the reactor. Near the wall, there was a net downward
flow of solids. This agrees well with the literature [16, 17, 18].
Table 1 compares the upward solids flux of the present study with the values measured by
Large et al. [19] with silica sand (dp > 250 µm) at ambient temperature and Ug = 30 cm/s.
Sampling time (s)
0 20 40 60 80 100 120 140 160 180
YC
O2 (vol%
)
0
1
2
3
4
5
6
7
z = 87 cm
137
Large et al. [19] measured solids entrainment 50 cm above the fluid bed and used extrapolation to
estimate the entrainment at the surface. Wen et al. [1] used these results to develop a correlation
for the entrainment at the bed surface. It can be seen from Table 1 that using the correlation of
Wen et al. [1] to estimate the entrainment at the bed surface yield a significant error: the present
study measured 17.6 kg/m2s and the correlation estimated 8.18×10-4 kg/m2s. Furthermore, the
measurements of Large et al. [19] were one order of magnitude smaller than the values obtained
in the present study. This difference can be explained by the extrapolation, which resulted in
significant errors.
Figure 5: Solids flux radial profile
Defining the location of the bed surface
The freeboard region begins at the fluidized bed surface. Bubbles form at the distributor, rise
rapidly through the bed and collapse at its surface releasing solid particle upward. The bubble
dynamics result in successive expansion and collapse of the fluidized bed. Depending on the gas
superficial velocity, solids entrainment can be significant and the bed surface can be diffuse.
In the present study, a time-average position of the bed surface was determined by pressure and
solids flux measurements. During the experiments, the static pressure was continually measured
at 4 axial positions at the wall and the time-averaged value was calculated. Figure 6 shows the
time-averaged static pressure inside the reactor. The pressure decreased significantly from
r (cm)
0 2 4 6 8 10
GSU
, D (kg/m
2.s
)
0
5
10
15
20
25
30
35
r (cm)
0 2 4 6 8 10
GS (kg/m
2.s
)
-12
-8
-4
0
4
8
12
29
35
45
z (cm)
(a) (b)
138
2.6 kPa to 0.3 kPa from z = 5 to 25 cm. These results suggest that these two probes were inside
the fluidized bed where the high solids fraction results in a significant hydrostatic pressure. In the
freeboard where the solids fraction is much lower, pressure remained constant at 0.1 kPa from
z = 35 to 77 cm. Therefore the pressure profile indicates that the bed surface was located between
z = 25 and 35 cm.
Table 1: Entrainment at the bed surface
Solids
type
Ug
(cm/s)
dp
(µµµµm)
Tb (°C) GSUo
(kg/m2s)
Reference
Silica sand 24 250 650 17.6 This work
Silica sand 24 250 650 8.18×10-4 Correlation - Wen et al. [1]
Silica sand 30 130 Ambient 2.2-3 Experimental data –
Large et al. [19] Silica sand 20 130 Ambient 10-2
The measurements of upward solids flux (or entrainment flux) were also used to determine the
bed surface position. Entrainment flux was measured at 5 axial positions (z = 29, 35, 46, 56 and
88 cm) and 6 radial positions (r = 0, 2, 4, 6, 8, 10 cm). Figure 6 shows the cross-sectional average
of entrainment flux as a function of axial position. The entrainment flux reached 17.6 kg/m2s at a
height of 29 cm and then decreased exponentially to 4.7 kg/m2s at z = 35 cm and zero at z = 46
cm. If the inflection point of the solid flux curve is taken as the bed surface as proposed by
Kunii et al. [20], the bed surface was located at z = 29 cm. This agrees well with the pressure
measurements. These results show that the position of the bed surface can be determined more
precisely with solids flux measurements compared to pressure measurements.
Species volume fraction in the freeboard
139
The species volume fractions were measured simultaneously to the solids flux at several positions
inside the freeboard. Figure 7 shows the radial profile of gaseous species at four axial positions
inside the freeboard. At the bed surface (z = 29 cm), there was a radial profile of propane volume
fraction – propane volume fraction decreased from 0.9 vol% at the centerline to 0.4 vol%
at 5 mm from the wall. At z = 35 cm, the radial profile was more uniform with an average
volume fraction of 0.3 vol%.
Figure 6: Defining the bed height
The radial profile of oxygen in the freeboard is shown in Figure 7(b). The oxygen radial profile
was approximately flat throughout the freeboard.
Figure 7: Species radial profile at bed surface
z (cm)
0 10 20 30 40 50 60 70 80 90
P (kP
a)
0.0
0.5
1.0
1.5
2.0
2.5
3.0
GS
U (kg/m
2.s
)
0
2
4
6
8
10
12
14
16
18Pressure
GSU
Calculated by
Hb
Sparger
z)0.26exp(103.5 4 −×
r (cm)
0 2 4 6 8 10
YC
3H
8 (m
ol%
)
0.0
0.2
0.4
0.6
0.8
1.0
r (cm)
0 2 4 6 8 10
YO
2 (m
ol%
)
12
16
20
29
35
45
56
z (cm)
140
Freeboard combustion
As the unconverted propane entered the freeboard, high temperatures combined to low solids
fraction resulted in propane combustion. Figures 8(a) and 8(b) show the axial profiles of C3H8,
CO, O2 and CO2 inside the fluidized bed reactor – propane conversion inside the fluid bed
reached 71% as propane decreased from 1.77 vol% (mixed-cup) at the inlet to 0.51 vol% at the
bed surface (z = 29 cm). In the freeboard, the remaining propane reacted with oxygen to form
CO2 and complete combustion was obtained at z = 56 cm.
Figure 8: Freeboard combustion
Temperature profile
Measuring the axial temperature profile inside the fluid bed reactor that corresponds to the
measured solids flux and species volume at each sample point is a difficult task unless the system
z (cm)
0 10 20 30 40 50 60 70 80 90
Y (m
ol%
)
0
3
6
9
12
15
18
21O
2
CO2
z (cm)
0 10 20 30 40 50 60 70 80 90
Y (m
ol%
)
0.0
0.3
0.6
0.9
1.2
1.5
1.8C
3H8
CO
(a)
(b)
Hb
Sparger
141
reaches steady. The temperature profile inside the freeboard was monitored with eight
thermocouples positioned along the length of the reactor. Two thermocouples were located inside
the fluidized bed and six in the freeboard. Figure 9 shows the axial temperature profiles just prior
to propane injection (t = 0 s) and after injection when one sample point was done (t ≈ 170 s).
After injection, the temperature inside the fluidized bed and freeboard increased with time due to
the exothermicity of the combustion reaction. Between each sample points, the propane was
stopped to allow the fluidized bed and freeboard to cool down. During combustion, a maximum
temperature increase of 33ºC was measured inside the freeboard. Using the chemical composition
of the gas measured at the bed surface (z = 29 cm) and assuming adiabatic constant pressure
combustion, the theoretical temperature increase is calculated as 546ºC from the CEA
equilibrium code [21].
Figure 9: Axial temperature profile
Figure 9 shows the variation in the axial temperature profiles (before and after propane injection)
for the entire experimental run (time required to sample from all axial and radial positions) - the
maximum temperature variation was ± 5ºC.
It was seen earlier from Figure 4 that only 20 seconds of sampling time was required to measure
species volume fractions. Figure 10 shows the temperature history at all sample points following
the injection of propane at t = 0 s. After the sampling time of approximately 170 s, the
temperature inside the reactor increased by a maximum of 2ºC within 20 seconds. Therefore, the
z (cm)
10 20 30 40 50 60 70 80 90 100
T (
oC
)
450
500
550
600
650
700
Prior to propane injection
After propane injection
∆TMAX = 33oCH
b
142
measured reactor temperature profile “after propane injection” corresponds to the measured
solids flux and species volume fractions.
Figure 10: Temperature history
CONCLUSIONS
In the present study, propane was injected inside a high temperature fluidized bed through a
downward facing sparger and the freeboard region was characterized using pressure, solids flux,
species volume fraction and temperature measurements during freeboard combustion. The
combustion in the fluidized bed and freeboard was confirmed by a decrease in oxygen and
propane as well as a temperature increase in the reactor. Solids flux and species volume fractions
were measured simultaneously using a non-isokinetic probe. The measurements showed a net
upward flow of solids at the centre of the reactor and a net downward flow near the wall.
Furthermore, the upward solids flux was found to decrease exponentially with height from the
bed surface. The measured entrainment at the bed surface was found to be significantly higher
than predicted by the correlation of Wen et al. [1]: 17.6 kg/m2s compared to 8.18×10-4 kg/m2s.
Pressure and solids flux measurements were also used to locate the fluidized bed surface. Solids
flux measurements were found to be more precise than pressure to determine bed height. The
location of the bed surface was taken as the inflection point of the entrainment curve. Finally, the
temperature profile corresponding to the measured solids flux and species volume fractions was
determined using a series of thermocouples positioned along the length of the reactor.
Time (s)
0 30 60 90 120 150 180
T (
oC
)
600
610
620
630
640
650
660
670z (cm)
19
25
40
51
56
Propane injection
143
NOMENCLATURE
A Cross-sectional area at the tip of the sampling probe (m2)
dp Average particle size (µm)
GS Net solids flux (kg/m2.s)
GSD Downward solids flux (kg/m2.s)
GSU Upward solids flux (kg/m2.s)
Hb Bed height (cm)
m Mass of sampled solids (kg)
r Radial position (cm)
t Time (s)
T Temperature (ºC)
Tb Fluidized bed temperature (ºC)
Q Sampling rate (mL/min)
Y Molar fraction (mol%)
z Axial position (cm)
REFERENCES
1. Wen, C.Y. & Chen, L.H. 1982. Fluidized bed freeboard phenomena: entrainment and
elutriation. AIChE J. 28 (1): 117-128.
2. Malleswara Rao, T.V. & Deo, G. 2007. Ethane and propane oxidation over supported
V2O5/TiO2 catalysts: analysis of kinetic parameters. Ind. Eng. Chem. Res. 46 (1): 70-79
3. Patience, G.S. & Lorences, M.J. 2006. VPO transient oxidation kinetics. Int. J. Chem. React.
Eng. 4 (A22).
144
4. Botella, P., Lopez Nieto, J.M. & Solsona, B. 2002. Selective oxidation of propene to acrolein
on Mo-Te mixed oxides catalyst prepared from ammonium telluromolybdates. J. Mol. Catal.
A-Chem. 184: 335-347.
5. Van der Vaart, D.R. 1988. Freeboard ignition of premixed hydrocarbon gas in a fluidised
bed. Fuel, 67: 1003-1007.
6. Hesketh, R.P. & Davidson, J.F. 1991. Combustion of methane and propane in an incipiently
fluidized bed. Combust. Flame. 85: 449-467.
7. Marinov, N.M., Pitz, W.J., Westbrook, C.K., Vincitore, A.M., Castaldi, M.J., Senkan, S.M.
and Melius, C.F. 1998. Aromatic and polycyclic aromatic hydrocarbon formation in a
laminar premixed n-Butane flame. Combust. Flame. 114 (1-2): 192-213.
8. Chen, L.H. & Wen, C.Y. 1982. Model of solid gas reaction phenomena in the fluidized bed
freeboard. AIChE J. 28 (6): 1019-1027.
9. Dounit, S., Hemati, M. & Steinmetz, D. 2001. Natural gas combustion in fluidized bed
reactors between 600 and 850ºC: experimental study and modelling of the freeboard. Powder
Technol. 120: 49-54.
10. Dounit, S., Hemati, M. & Andreux, R. 2008. Modelling and experimental validation of a
fluidized-bed reactor freeboard region: Application to natural gas combustion. Chem. Eng. J.
140: 457-465.
11. Rhodes, M.J. & Laussmann, P. 1992. A simple non-isokinetic sampling probe for dense
suspensions. Powder Technol. 70: 141-151.
12. Reinhardt, B., Cordonnier, A. & Florent, P. 1999. Use of an isokinetic sampling propane.
Results in a cyclone. Powder Technol. 101: 81-90.
13. Van der Meer, E.H., Thorpe, R.B. & Davidson, J.F. 2000. Flow patterns in the square cross-
section riser of a circulating fluidized bed and the effect of riser exit design. Chem. Eng. Sci.
55: 4079-4099.
14. Morooka S., Kawazuishi, K. & Kato, Y. 1980. Holdup and flow pattern of solids particles in
freeboard of gas-solid fluidized bed with fine particles. Powder Technol. 26: 75-82.
15. Werther, J. 1999. Measurement techniques in fluidized beds. Powder Technol. 102: 15-36.
16. Aguillón, J., Shakourzadeh, K. & Guigon, P. 1995. Comparative study of non-isokinetic
sampling probes for solids flux measurement in circulating fluidized beds. Powder Technol.
83: 79-84.
145
17. Van Breugel, J.W., Stein, J.J.M. & de Vries, R.J. 1970. Isokinetic sampling in a dense gas-
solids stream. Proc. Instn. Mech. Engrs. 184: 18-23.
18. Rhodes, M.J., Laussmann, P., Villain, F. & Geldart, D. 1988. Measurement of radial and
axial solids flux variations in the riser of a circulating fluidized bed. In Basu, P. & Large, J.F.
(eds) Circulating Fluidized Bed Technology II: 155-164. Oxford: Pergamon Press.
19. Large J.F., Martinie, Y. & Bergougnou, M.A. 1977. Interpretative model for entrainment in a
large gas-fluidized bed. J. Powders Bulk Solids Technol. 1: 15-21.
20. Kunii, D. & Levenspiel, O. 1991. Fluidization Engineering. 2nd edition. Boston, MA:
Butterworth-Heinemann.
21. Gordon S. & McBride, B.J. 1994. Computer program for calculation of complex chemical
equilibrium compositions and applications. NASA RP-1311.
146
ANNEXE 3 – ARTICLE PRÉSENTÉ À LA CONFÉRENCE FBC 2009
(CHINE)
GAS-PHASE COMBUSTION IN THE FREEBOARD OF A FLUIDIZED BED –
FREEBOARD CHARACTERIZATION
Jean-Philippe Laviolette, Gregory S. Patience and Jamal Chaouki
Department of Chemical Engineering, Ecole Polytechnique, Montréal, Canada
Abstract: The prediction of propane autoignition in the freeboard of a fluidized bed is
complicated by the presence of solids, intermediate products and non-homogeneities
(solids, temperature and species gradients) that should be accounted for in a reaction model.
However, the simultaneous characterization of these parameters during combustion is very
challenging. An experimental study of propane combustion inside the freeboard
(I.D. = 0.2 m) of a fluidized bed of sand (dp = 290 µm) was performed at a low superficial
gas velocity (Ug = 0.24 m/s). Propane was injected inside the fluidized bed (TBed = 650ºC)
through a downward-facing sparger. Also, solids flux and species volume fractions were
measured using a non-isokinetic sampling probe. The results showed an exponential
decrease with height of the upward solids flux (GSU) – GSU was zero at 0.17 m above the
bed surface, which was taken as the inflection point of the GSU curve. GSUo measurements
were significantly higher than the values given by the correlation of Wen et al. (1982). The
bed surface (boundary condition) and freeboard were characterized by measuring pressure,
solids flux, species volume fractions and temperature at several radial and axial positions.
During the experiments, the fluidized bed achieved a pseudo steady-state operation that
ensured that the measured temperature profile corresponded to the solids flux and species
fractions. Partial propane combustion in the fluidized bed (71%) produced CO and cracking
species that were transported in the freeboard. Complete combustion occurred within
0.15 m of the bed surface and the propane induction time in the freeboard (< 0.25 s) was on
the same order as the values given by three induction time correlations for homogeneous
systems.
147
Keywords: propane, freeboard, induction time, non-isokinetic sampling, combustion
INTRODUCTION
Homogeneous reactions generally result in yield lost of the desired compound, but they may also represent
a significant safety hazard in the case of fast and highly exothermic reactions such as those involving
oxygen and hydrocarbons. Chemical processes with oxygen and hydrocarbons are widespread in the
industry and new processes are currently under development in the context of oil price fluctuations and
global warming: selective oxidation of alkanes, combustion of cheaper feedstock for heating and catalyst
regeneration, biomass gasification, etc. Fluidized bed reactors are currently being developed for a variety
of these processes. In order to maximize productivity, they may operate within the explosion limits while
feeding the oxidant and hydrocarbon separately into the bed – through spargers. This is possible due to
the ability of the solids phase to suppress the homogeneous reactions (combustion) and at the same time
promote selective heterogeneous reactions. However, downstream of the solids – in the freeboard,
cyclones and associated piping – the effluent gas phase composition is potentially explosive – high
hydrocarbon and oxygen concentrations, elevated temperature with an insufficient solids volume fraction
to quench non-selective homogeneous reactions (Van der Vaart, 1988; Hesketh et al., 1991). Minimizing
the risk of gas-phase combustion-deflagration in these regions remains an important design issue.
Several explosion criteria (explosion limits, autoignition temperature and induction times
(Hesketh et al., 1991)) measured in homogeneous gas-phase systems as well as gas-phase kinetic models
are available in the scientific literature (Marinov et al., 1998). Nevertheless, extrapolating these tendencies
to a fluidized bed reactor introduces uncertainties since the entrained solid particles affect the gas-phase
reactions downstream of the fluidized bed. Furthermore, fluidized beds are non-homogeneous systems -
solids entrainment decreases exponentially from the bed surface to the transport disengagement height
(Wen et al., 1982). Also, an axial temperature gradient may exist in the freeboard region as well as species
concentration gradients depending on the hydrodynamics.
To accurately predict gas-phase combustion in the freeboard of a fluidized bed for a wide range of
operational conditions, a freeboard combustion model that combines the gas/solids hydrodynamics and the
reaction kinetics must be developed. This model needs to integrate through the axial gradient of solids
fraction and temperature that characterize the freeboard as well as accounting for the effects of solids on
the reaction kinetics. Furthermore, the model should account for the partial combustion and cracking
148
product originating from the fluidized bed. On the other hand, the simultaneous characterization of solids
fraction, species concentration, temperature and pressure in the bed and freevoard during combustion is
very challenging.
Freeboard reaction models have been previously proposed by Chen et al. (1982), Dounit et al. (2001) and
Dounit et al. (2008). To estimate solids entrainment, the correlation of Wen et al. (1982) was used, which
is based on the extrapolation of data measured several centimetres downstream of the bed surface
(sometimes more than 0.5 m). Consequently, the existing entrainment models may yield significant errors
near the bed surface and better accuracy would be obtained by directly measuring solids fraction. Also, the
location of the bed surface (beginning of the freeboard region) and its corresponding boundary condition
(solids fraction/flux, species fractions, temperature and pressure) need to be clearly defined.
In the present study, propane was injected inside a high temperature fluidized bed through a downward
facing sparger and the freeboard region was characterized by simultaneously measuring pressure, solids
flux, species volume fraction and temperature during freeboard combustion. Solids flux and species
volume fractions were measured using a non-isokinetic probe. The temperature profile in the freeboard
was monitored using eight thermocouples.
EXPERIMENTAL APPARATUS
The experiments were conducted in the fluidized bed reactor of Figure 1. The bed and freeboard regions
had an inner diameter of 20 cm. For all experiments, sand particles (Geldart group B, ρs = 2650 kg/m3 and
dp = 290 µm) were used as the bed material and air was the fluidizing media. The experimental procedure
first consisted of adjusting the superficial gas velocity to 0.24 m/s (at 650oC). The fluidized bed was then
heated with the natural gas burner upstream of the windbox and by injecting heavy vacuum gas oil
(HVGO) through the sparger. When the bed temperature reached 680oC, the burner and HVGO injection
were stopped and propane was injected through the sparger for 30 minutes to burn HVGO residues.
The propane was turned off and the fluidized bed cooled down until the bed temperature reached 645oC.
Propane was then injected through the downward sparger to obtain a mixed-cup volume fraction of
2 vol% in air. A non-isokinetic probe was used to sample gas and solids simultaneous from 5 axial and 6
radial positions. The sampling system consisted of a 4.1 mm I.D. probe bent at a 90o at its tip. The
149
sampled gas/solid stream first went through a 250 mL container that lowered the sampled gas velocity and
caused the separation of the solid particles from the gas. The gas was analyzed by a Siemens Ultramat
CO/CO2 analyzer and a Varian CP-4900 micro-GC that measured the following species: O2, H2, CH4,
C2H4, C2H6, C3H6, C3H8, CO, CO2 and N2. Solids flux was calculated from the mass of the sampled solids
and the sampling time.
Fig. 1: Fluidized bed reactor
The upward (GSU) and downward (GSD) solid fluxes were measured by orienting the probe tip downward
and upward, respectively. The sampling flow rate was controlled using a rotameter connected to a vacuum
line. Gas was also continuously sampled in the disengagement region (z = 1.47 m) and analyzed with an
oxygen analyzer (Siemens Oxymat 6.1). The results from the GC, CO/CO2 analyzer and oxygen analyzer
were compared and used in a molar balance to confirm the validity of the measured species volume
fractions. The temperature along the reactor was monitored using 8 thermocouples located along the
length of the reactor. The sparger consisted of a 4.1 mm I.D. tube pointing downward at the bed
centreline. The sparger tip was located at a height of 0.098 m above the distributor.
RESULTS AND DISCUSSION
3 4
Non-isokinetic sampling probe
0.48 m
0.2 m
0.97 m
0.38 m
1 1st cyclone
2 2nd
cyclone 3 Natural gas burner 4 Windbox 5 Fluidized bed
0.6 m
0.2 m
Air + Natural gas
C3H8
r
5
O2 analyser
CO/CO2 analyser
z
1
2
Gas chromatograph
150
The reaction rate of propane combustion in the freeboard of a fluid bed depends on species volume
fractions, temperature and solids fraction. They are dependent parameters and should be measured
simultaneously.
Freeboard characterization
Simultaneous sampling of gas and solids
A non-isokinetic sampling probe was used to measure the species volume fractions and the solids flux
simultaneously. A preliminary set of experiments showed that the measured solids flux was almost
independent of sampling rate in the range of 270 - 470 mL/min. Therefore, it was not necessary to have
sampling that was precisely iso-kinetic and a sampling rate of 425 mL/min was used for all experiments.
Results also showed that a minimum sampling time of 60 seconds was required to reach a permanent
condition. The gas sampled with the non-isokinetic sampling probes was sent to the CO/CO2 analyzer and
the GC for analysis. A sampling time of only 20 seconds was required for a representative measurement of
the species volume fractions inside the reactor.
Fig. 2: Solids flux radial profile
Solids flux in the freeboard
Upward (GSU) and downward (GSD) solid fluxes were measured at 5 axial positions and 6 radial positions.
The radial profiles of solids flux are shown in Figure 2(a) for three axial positions: 0.29, 0.35 and 0.45 m
above the distributor. The upward (empty symbols) and downward (filled symbols) solids fluxes reached a
maximum at the reactor centreline and decreased closer to the wall. The measurements were averaged
over the fluid bed cross-section. At Z = 0.29 m, the upward and downward solids fluxes were 17.6 kg/m2s
and 19.0 kg/m2s, respectively (difference of 1.4 kg/m2s). At Z = 0.35 m, the upward solids flux was
r (cm)
0 2 4 6 8 10
GS
U, D (kg/m
2.s
)
0
5
10
15
20
25
30
35
r (cm)
0 2 4 6 8 10
GS (kg/m
2.s
)
-12
-8
-4
0
4
8
12
29
35
45
z (cm)
(a) (b)
151
ZF (m)
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
GS
U (
kg
/m2s)
10-5
10-4
10-3
10-2
10-1
100
101
102
This work
Data of Large et al. (1976)
FZSU eG 2618 −=
( ) FZSU eG 43104 −−×=
Z (m)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
P (
kP
a)
0.0
0.5
1.0
1.5
2.0
2.5
3.0
GS
U (
kg
/m2s)
0
2
4
6
8
10
12
14
16
18Pressure
Upward GSHB
4.7 kg/m2s and the downward solids flux was 3.0 kg/m2s (difference of 1.7 kg/m2s). At Z = 0.45 m, the
upward and downward solids flux were both zero. The radial profiles of net solids flux (GS) are shown in
Figure 2(b) for three axial positions: there was a net upward of solids at the centre of the reactor. Near the
wall, there was a net downward flow of solids. This agrees well with the literature (Aguillón et al., 1995;
Van Breugel et al., 1970; Rhodes et al., 1988).
At each axial position, the measurements were averaged over the fluidized bed cross-section. Figure 3(a)
compares the upward solids flux of the present study with the values given by the correlation of
Wen et al. (1982). The correlation of Wen et al. (1982) underestimated GSU at the bed surface: the present
study measured 17.6 kg/m2s and the correlation estimated 8.18×10-4 kg/m2s. Wen et al. (1982) derived the
correlation by extrapolating measurements of Large et al. (1977) made at 0.50 m above the bed surface,
which may explain the observed discrepancy.
Defining the bed surface
In the present study, a time-average position of the bed surface was determined by pressure and solids flux
measurements. Figure 3(b) shows the time-averaged static pressure measured at 4 axial positions at the
wall and the cross-sectional mean of GSU measured at 5 axial positions and 6 radial positions. The pressure
decreased significantly in the bed from 2.6 kPa to 0.3 kPa (Z = 0.05 to 0.25 m) and remained constant at
0.1 kPa (Z = 0.35 to 0.77 m) in the freeboard. This indicates that the bed surface was located between
Z = 0.25 and 0.35 m.
Fig. 3: Upward solids flux and pressure measurements
(a) (b)
Sparger
152
51
56
Time (s)
0 30 60 90 120 150 180
T (
oC
)
600
610
620
630
640
650
660
670
Z (m)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
T (
oC
)
450
500
550
600
650
700
Air
Air + C3H8
∆TMAX = 33oC Z (m)
0.19
0.25
0.40
0.51
0.56
HB
Propane injection Sparger
(a) (b)
The upward solids flux reached 17.6 kg/m2s at a height of 0.29 m and then decreased exponentially to
4.7 kg/m2s at Z = 0.35 m and zero at Z = 0.46 m. If the inflection point of the solid flux curve is taken as
the bed surface as proposed by Kunii et al. (1991), the bed height (HB) was 0.29 m. This agrees well with
the pressure measurements.
Temperature profile
The temperature profile inside the freeboard was monitored with eight thermocouples positioned along the
length of the reactor. Two thermocouples were located inside the fluidized bed and six in the freeboard.
Figure 4(a) shows the axial temperature profiles just prior to propane injection (t = 0 s) and after injection
when one sample point was done (t ≈ 170 s). After injection, the temperature inside the fluidized bed and
freeboard increased with time due to the exothermicity of the combustion reactions. Between each sample
points, the propane was stopped to cool down the fluidized bed and freeboard. During combustion, a
maximum temperature increase of 33ºC was observed inside the freeboard. Using the chemical
composition of the gas measured at the bed surface (z = 0.29 m) and assuming adiabatic constant pressure
combustion, the theoretical temperature increase is calculated as 546ºC from the CEA equilibrium
code (Gordon et al., 1994). The temperature in the freeboard reached a maximum of 670ºC during the
experiments.
Fig. 4: Freeboard temperature profile
Figure 4(b) shows the temperature history of 5 thermocouples in the freeboard following the injection of
propane at t = 0 s. After the sampling time of approximately 170 s, the temperature inside the reactor
increased by a maximum of 2ºC within 20 seconds, which was the required sampling time for a
representative measurement of the species volume fractions inside the reactor. Therefore, the measured
153
r (cm)
0 2 4 6 8 10
Y (
mo
l%)
0.0
0.2
0.4
0.6
0.8
1.0C
3H
8
CO
r (cm)
0 2 4 6 8 10
Y (
mol%
)
0.00
0.05
0.10
0.15
0.20H
2
CH4
C2H
4
C2H6
C3H
6
(a) (b)
solids flux and species volume fractions corresponded to the temperature profile with an uncertainty of
±2ºC.
Figure 5: Species at the bed surface
Species at the bed surface
Partial combustion of C3H8 inside the fluidized bed produced CO and CO2. Furthermore, the high
temperature resulted in C3H8 cracking. Figures 5(a) and 5(b) shows the volume fractions of C3H8, CO, H2,
CH4, C2H4, C2H6 and C3H6 at the bed surface (Z = 0.29 cm). The volume fraction of C3H8 decreased from
0.9% at the centreline to 0.4% at 5 mm from the wall. The volume fractions of CO, H2, CH4, C2H4, C2H6
and C3H6 were approximately constant throughout the reactor radius at 0.45%, 0.07%, 0.01%, 0.06%,
0.04% and 0.04%, respectively.
Figure 6: Freeboard combustion
Z (m)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Y (
mo
l%)
0
3
6
9
12
15
18
21O2
CO2
Z (m)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Y (
mo
l%)
0.0
0.3
0.6
0.9
1.2
1.5
1.8C3H8
CO
(a)HB
Sparger
(b)
154
Freeboard combustion
Figures 6(a) and 6(b) show the axial profiles (cross-sectional average) of C3H8, CO, O2 and CO2 inside the
fluidized bed reactor – propane conversion inside the fluid bed reached 71% as propane decreased from
1.77 vol% (mixed-cup) at the inlet to 0.51 vol% at the bed surface (Z = 0.29 m). In the freeboard, the
remaining propane reacted with oxygen to form CO2 and complete combustion was obtained at
Z = 0.56 m (17 cm above the bed surface).
Table 1 compares the measured induction time with the values from three propane induction time
correlations for homogeneous systems that are available in the scientific literature. The induction time is
given for the bed temperature (650ºC) and the maximum temperature observed inside the freeboard
(670ºC). Also, the C3H8 and O2 at the bed surface were used. In the present work, the measured induction
time was below 0.25 s, which was the gas residence time between the bed surface (Z = 0.29 m) and the
next sampling point (Z = 0.35 m). Between these two sampling positions, a 53% decrease in propane was
observed. The measured induction time was on the same order as the values given by the correlations.
Table 1: Induction time data
ττττ (s) Reference
650ºC 670ºC
0.28 0.18 Burcat et al., 1971
0.07 0.06 Penyazkov et al., 2005
0.68 0.33 Kim et al., 2001
< 0.25 This work
CONCLUSIONS
An experimental study of propane combustion inside the freeboard (I.D. = 0.2 m) of a fluidized bed of
sand (dp = 290 µm) was performed at low superficial gas velocity (Ug = 0.24 m/s). Propane was injected
155
inside the fluidized bed (TBed = 650ºC) through a downward-facing sparger. Solids flux and species
volume fractions were measured using a non-isokinetic sampling probe. The results showed an
exponential decrease with height of the upward solids flux (GSU) – GSU was zero at 0.17 m above the bed
surface, which was taken as the inflection point of the GSU curve. GSUo measurements were significantly
higher than the values given by the correlation of Wen et al. (1982). The bed surface (boundary condition)
and freeboard were characterized by measuring pressure, solids flux, species volume fractions and
temperature at several radial and axial positions. During the experiments, the fluidized bed achieved a
pseudo steady-state operation that ensured that the measured temperature profile corresponded to the
solids flux and species fractions. Partial propane combustion in the fluidized bed (71%) produced CO and
cracking species that were transported in the freeboard. Complete combustion occurred within 0.15 m of
the bed surface and the propane induction time in the freeboard (< 0.25 s) was on the same order as the
values given by three induction time correlations for homogeneous systems.
References
Aguillón, J., Shakourzadeh, K., Guigon, P.: Powder Technol. 83 (1995), pp. 79-84.
Botella, P., Lopez Nieto, J.M., Solsona, B.: J. Mol. Catal. A-Chem. 184 (2002), pp. 335-347.
Burcat, A., Lifshitz, A.: Proceedings of the 13th International Symposium on Combustion, Combustion
Institute (1971), pp. 745-755.
Chen, L.H.,Wen, C.Y.: AIChE J. 28 (6) (1982), pp. 1019-1027.
Dounit, S., Hemati, M., Steinmetz, D.: Powder Technol. 120 (2001), pp. 49-54.
Dounit, S., Hemati, M., Andreux, R. Chem. Eng. J. 140 (2008), pp. 457-465.
Gordon S., McBride, B.J.: Computer program for calculation of complex chemical equilibrium
compositions and applications. NASA RP-1311 (1994).
Hesketh, R.P., Davidson, J.F.: Combust. Flame 85 (1991), pp. 449-467.
Kunii, D., Levenspiel, O.: Fluidization Engineering, Butterworth-Heinemann, 1991.
Kim, K., Soo Shin, K.: Bull. Korean Chem. Soc. 22 (3) (2001), pp. 303-307.
Large J.F., Martinie, Y., Bergougnou, M.A.: J. Powders Bulk Solids Technol. 1 (1977), pp. 15-21.
Malleswara Rao, T.V., Deo, G.: Ind. Eng. Chem. Res. 46 (1) (2007), pp. 70-79.
Marinov, N.M., Pitz, W.J., Westbrook, C.K., Vincitore, A.M., Castaldi, M.J., Senkan, S.M., Melius, C.F.:
Combust. Flame. 114 (1-2) (1998), pp. 192-213.
Morooka S., Kawazuishi, K., Kato, Y.: Powder Technol. 26 (1980), pp. 75-82.
Patience, G.S., Lorences, M.J.: Int. J. Chem. React. Eng. 4 (A22) (2006).
156
Penyazkov, O.G., Ragotner, K.A., Dean, A.J., Varatharajan, B.: Proceedings of the Combustion Institute,
Elsevier Inc. (2005) pp. 1941-1947.
Reinhardt, B., Cordonnier, A., Florent, P.: Powder Technol. 101 (1999), pp. 81-90.
Rhodes, M.J., Laussmann, P., Villain, F., Geldart, D.: Circulating Fluidized Bed Technology II (Basu, P.,
Large, J.F. (eds)), Pergamon Press Oxford (1988), pp. 155-164.
Rhodes, M.J., Laussmann, P.: Powder Technol. 70 (1992), pp. 141-151.
Van Breugel, J.W., Stein, J.J.M., de Vries, R.J.: Proc. Instn. Mech. Engrs. 184 (1970), pp. 18-23.
Van der Vaart, D.R.: Fuel 67 (1988), pp. 1003-1007.
Van der Meer, E.H., Thorpe, R.B., Davidson, J.F.: Chem. Eng. Sci. 55 (2000), pp. 4079-4099.
Wen, C.Y., Chen, L.H.: AIChE J. 28 (1) (1982), pp. 117-128.
Werther, J.: Powder Technol. 102 (1999), pp. 15-36.
Notation
Symbols
dp average particle size µm
GS solids flux kg/m2s
r radial position cm
t time s
T temperature ºC
Ug superficial gas velocity cm/s
Z axial position m
Greek letters
τ induction time s
Subscripts
Bed fluidized bed
D downwards
F freeboard
g gas phase
o fluidized bed surface
s solid phase
U upwards
157
ANNEXE 4 – ARTICLE PRESENTÉ À LA CONFÉRENCE WCCE8 2009
(MONTRÉAL, CANADA)
GAS-PHASE COMBUSTION IN THE FREEBOARD OF A FLUIDIZED BED
Jean-Philippe Laviolette, Gregory S. Patience and Jamal Chaouki
Ecole Polytechnique, Montréal, Canada
Department of Chemical Engineering
Abstract: The prediction of propane autoignition in the freeboard of a fluidized bed is
complicated by the presence of solids, intermediate products and non-homogeneities (solids,
temperature and species gradients). An experimental and modeling study of propane
combustion inside a fluidized bed and the freeboard (I.D. = 20 cm) was performed. Propane
was injected inside a bubbling fluidized bed (TBed = 650ºC and Ug = 24 cm/s) of sand
particles (dp = 290 µm) through a downward-facing sparger. The combustion process in the
bed and freeboard was characterized by measuring pressure, solids flux, species volume
fractions and temperature at several radial and axial positions. The results were compared to a
homogeneous microkinetic model and induction time correlations from the literature. The
results showed that the upward solids flux (GSU) decreased exponentially with height and the
bed surface was defined as the inflection point. GSU measurements were significantly higher
than the values given by the correlation of Wen et al. (1982). GS was zero at a distance of
15 cm above the bed and the combustion was complete within 27 cm. The measured
induction time was on the same order as the values given by three induction time correlations.
However, reaction rates from a homogeneous microkinetic model were significantly lower
than the experimental values.
Keywords: fluidized bed, propane, combustion, freeboard, non-isokinetic sampling, solids
flux, induction time, microkinetic modelling
1. INTRODUCTION
158
Homogeneous reactions generally result in yield lost of the desired compound, but they may also represent a
significant safety hazard in the case of fast and highly exothermic reactions such as those involving oxygen and
hydrocarbons. Chemical processes with oxygen and hydrocarbons are widespread in the industry and new processes
are currently under development in the context of oil price fluctuations and global warming: selective oxidation of
alkanes, combustion of cheaper feedstock for heating and catalyst regeneration, biomass gasification, etc. Fluidized
bed reactors are currently being developed for a variety of these processes. In order to maximize productivity, they
may operate within the explosion limits while feeding the oxidant and hydrocarbon separately into the bed – through
spargers. This is possible due to the ability of the solids phase to suppress the homogeneous reactions (combustion)
and at the same time promote selective heterogeneous reactions. However, downstream of the solids – in the
freeboard, cyclones and associated piping – the effluent gas phase composition is potentially explosive – high
hydrocarbon and oxygen concentrations, elevated temperature with an insufficient solids volume fraction to quench
non-selective homogeneous reactions (Van der Vaart, 1988; Hesketh et al., 1991). Minimizing the risk of gas-phase
combustion-deflagration in these regions remains an important design issue.
Several explosion criteria (explosion limits, autoignition temperature and induction times (Hesketh et al., 1991))
measured in homogeneous gas-phase systems as well as gas-phase kinetic models are available in the scientific
literature (Marinov et al., 1998). However, extrapolating these tendencies to a fluidized bed reactor introduces
uncertainties since the entrained solid particles affect the gas-phase reactions downstream of the fluidized bed.
Furthermore, fluidized beds are non-homogeneous systems - solids entrainment decreases exponentially from the bed
surface to the transport disengagement height (Wen et al., 1982). Also, an axial temperature gradient may exist in the
freeboard region as well as species concentration gradients depending on the hydrodynamics.
To accurately predict gas-phase combustion in the freeboard of a fluidized bed for a wide range of operational
conditions, a freeboard combustion model that combines the gas/solids hydrodynamics and the reaction kinetics must
be developed. This model needs to integrate through the axial gradient of solids fraction and temperature that
characterize the freeboard as well as accounting for the effects of solids on the reaction kinetics. Models of the
reaction in the freeboard of fluidized beds have been previously proposed by Chen et al. (1982), Dounit et al. (2001)
and Dounit et al. (2008). These models used entrainment models that include a correlation for the entrainment at the
bed surface such as the model of Wen et al. (1982). However, the solids entrainment at the bed surface was
extrapolated from experimental data measured several centimetres downstream of the bed surface (sometimes more
than 50 centimetres). Therefore, the existing entrainment models may yield significant errors near the bed surface
and better accuracy would be obtained by directly measuring solids fraction.
In the present study, propane was injected inside a high temperature fluidized bed through a downward facing
sparger and the freeboard region was characterized using pressure, solids flux, species volume fraction and
159
temperature measurements during freeboard combustion. The fluidized bed temperature, superficial velocity and
propane/air ratio were adjusted to obtain propane combustion in the freeboard. Solids flux and species volume
fractions were measured simultaneously using a non-isokinetic probe. The temperature profile in the freeboard was
monitored using eight thermocouples.
2. EXPERIMENTAL APPARATUS
Fig. 1: Fluidized bed reactor
The experiments were conducted in the fluidized bed reactor of Figure 1. The bed and freeboard regions had an inner
diameter of 20 cm. For all experiments, sand particles (Geldart group B ρs = 2650 kg/m3, dp = 290 µm) were used as
the bed material and air was the fluidizing media. The experimental procedure first consisted of adjusting the
superficial gas velocity to 24 cm/s (at 650oC). The fluidized bed was then heated with the natural gas burner
upstream of the windbox and by injecting heavy vacuum gas oil (HVGO) through the sparger. When the bed
temperature reached 680oC, the burner and HVGO injection were stopped and propane was injected through the
sparger for 30 minutes to burn HVGO residues.
3 4
Non-isokinetic sampling probe
0.48 m
0.2 m
0.97 m
0.38 m
1 1st cyclone
2 2nd
cyclone 3 Natural gas burner 4 Windbox 5 Fluidized bed
0.6 m
0.2 m
Air + Natural gas
C3H8
r
5
O2 analyser
CO/CO2 analyser
z
1
2
Gas chromatograph
160
The propane was turned off and the fluidized bed cooled down until the bed temperature reached 645oC. Propane
was then injected through the downward sparger to obtain a mixed-cup volume fraction of 2 vol% in air. A non-
isokinetic probe was used to sample gas and solids simultaneous from 5 axial and 6 radial positions. The sampling
system consisted of a 4.1 mm I.D. probe bent at a 90o at its tip. The sampled gas/solid stream first went through a
250 mL container that lowered the sampled gas velocity and caused the separation of the solid particles from the gas.
The gas was analyzed by a Siemens Ultramat CO/CO2 analyzer and a Varian CP-4900 micro-GC that measured the
following species: O2, H2, CH4, C2H4, C2H6, C3H6, C3H8, CO, CO2 and N2. Solids flux was calculated from the mass
of the sampled solids and the sampling time:
The upward (GSU) and downward (GSD) solid fluxes were measured by orienting the probe tip downward and
upward, respectively. The sampling flow rate was controlled using a rotameter connected to a vacuum line. Gas was
also continuously sampled in the disengagement region (z = 147 cm) and analyzed with an oxygen analyzer
(Siemens Oxymat 6.1). The results from the GC, CO/CO2 analyzer and oxygen analyzer were compared and used in
a molar balance to confirm the validity of the measured species volume fractions. The temperature along the reactor
was monitored using 8 thermocouples located along the length of the reactor. The sparger consisted of a 4.1 mm I.D.
tube pointing downward at the bed centreline. The sparger tip was located at a height of 9.8 cm above the distributor.
3. RESULTS AND DISCUSSION
The reaction rate of propane combustion in the freeboard of a fluid bed depends on species volume fractions,
temperature and solids fraction. They are dependent parameters and should ideally be measured simultaneously.
3.1 Simultaneous sampling of gas and solids
A non-isokinetic sampling probe was used to measure the species volume fractions and the solids flux
simultaneously. A preliminary set of experiments showed that the measured solids flux was almost independent of
sampling rate in the range of 270 - 470 mL/min. Therefore, it was not necessary to have sampling that was precisely
iso-kinetic and a sampling rate of 425 mL/min was used for all experiments. Results also showed that a minimum
sampling time of 60 seconds was required to reach a permanent condition. The gas sampled with the non-isokinetic
sampling probes was sent to the CO/CO2 analyzer and the GC for analysis. A sampling time of only 20 seconds was
required for a representative measurement of the species volume fractions inside the reactor.
161
ZF (m)
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
GS
U (
kg/m
2s)
10-5
10-4
10-3
10-2
10-1
100
101
102
This work
Data of Large et al. (1976)
FZSU eG 2618 −=
( ) FZSU eG 43104 −−×=
3.2 Solids flux in the freeboard
Upward (GSU) and downward (GSD) solid fluxes were measured at 5 axial positions and 6 radial positions. At each
axial position, the measurements were averaged over the fluidized bed cross-section. The radial profiles of net solids
flux (GS) showed a net upward of solids at the centre of the reactor. Near the wall, there was a net downward flow of
solids. This agrees well with the literature (Aguillón et al., 1995; Van Breugel et al., 1970; Rhodes et al., 1988).
Figure 2 compares the upward solids flux of the present study with the values given by the correlation of
Wen et al. (1982). Figure 2 shows that the correlation of Wen et al. (1982) underestimates GSU at the bed surface: the
present study measured 17.6 kg/m2s and the correlation estimated 8.18×10-4 kg/m2s. Wen et al. (1982) derived the
correlation by extrapolating measurements of Large et al. (1977) made at 50 cm above the bed surface, which may
explain the observed discrepancy.
3.3 Defining the bed surface
In the present study, a time-average position of the bed surface was determined by pressure and solids flux
measurements. Figure 3 shows the time-averaged static pressure measured at 4 axial positions at the wall and the
cross-sectional mean of entrainment flux measured at 5 axial positions and 6 radial positions. The pressure decreased
significantly in the bed from 2.6 kPa to 0.3 kPa (Z = 5 to 25 cm) and remained constant at 0.1 kPa (Z = 35 to 77 cm)
in the freeboard. This indicates that the bed surface was located between Z = 25 and 35 cm.
Fig. 2: Upward solid flux
162
Fig. 3: Bed height
The entrainment flux reached 17.6 kg/m2s at a height of 29 cm and then decreased exponentially to 4.7 kg/m2s at
Z = 35 cm and zero at Z = 46 cm. If the inflection point of the solid flux curve is taken as the bed surface as proposed
by Kunii et al. (1991), the bed height (HB) was 29 cm. This agrees well with the pressure measurements.
3.4 Temperature profile
The temperature profile inside the freeboard was monitored with eight thermocouples positioned along the length of
the reactor. Two thermocouples were located inside the fluidized bed and six in the freeboard. Figure 4(a) shows the
axial temperature profiles just prior to propane injection (t = 0 s) and after injection when one sample point was done
(t ≈ 170 s). After injection, the temperature inside the fluidized bed and freeboard increased with time due to the
exothermicity of the combustion reactions. Between each sample points, the propane was stopped to allow the
fluidized bed and freeboard to cool down. During combustion, a maximum temperature increase of 33ºC was
measured inside the freeboard. Using the chemical composition of the gas measured at the bed surface (z = 29 cm)
and assuming adiabatic constant pressure combustion, the theoretical temperature increase is calculated as 546ºC
from the CEA equilibrium code (Gordon et al., 1994).
Figure 4(b) shows the temperature history at all sample points following the injection of propane at t = 0 s. After the
sampling time of approximately 170 s, the temperature inside the reactor increased by a maximum of 2ºC within 20
seconds (required sampling time to measure species with GC and Ultramat). Therefore, the reactor temperature
profile corresponds to the measured solids flux and species volume fractions.
Z (m)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
P (
kP
a)
0.0
0.5
1.0
1.5
2.0
2.5
3.0
GS
U (
kg/m
2s)
0
2
4
6
8
10
12
14
16
18Pressure
Upward GSHB
Sparger
163
Fig. 4: Freeboard temperature profile
3.5 Species fraction in the freeboard and freeboard combustion
The results showed a radial profile of species volume fraction at the bed surface (z = 29 cm) – propane volume
fraction decreased from 0.9 vol% at the centerline to 0.4 vol% at 5 mm from the wall. At z = 35 cm, the radial profile
was more uniform with an average volume fraction of 0.3 vol%. Figures 5(a) and 5(b) show the axial profiles (cross-
sectional average) of C3H8, CO, O2 and CO2 inside the fluidized bed reactor – propane conversion inside the fluid
bed reached 71% as propane decreased from 1.77 vol% (mixed-cup) at the inlet to 0.51 vol% at the bed surface
(z = 29 cm). In the freeboard, the remaining propane reacted with oxygen to form CO2 and complete combustion
was obtained at z = 56 cm.
Fig. 5: Freeboard combustion
Z (m)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Y (
mo
l%)
0
3
6
9
12
15
18
21O2
CO2
Z (m)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Y (
mo
l%)
0.0
0.3
0.6
0.9
1.2
1.5
1.8C3H8
CO
(a)HB
Sparger
(b)
164
Table 1 compares the measured induction time with the values given by three correlations available in the scientific
literature – the measured induction time is on the order of the values given by the correlations.
TABLE 1 Induction time data
T (ºC) ττττ (s) Reference
- < 0.25 This work
650 0.28 Burcat et al., 1971
650 0.07 Penyazkov et al., 2005
650 0.68 Kim et al., 2001
4. MODELLING
A homogeneous microkinetic model (Marinov et al., 1998) was used to model the freeboard combustion. The gas-
phase hydrodynamics was modeled as plug flow. Figure 6 compares the modelling and experimental results – the
model underestimates the combustion rates. This may be explained by the presence of free radicals that come from
partial combustion in the fluidized bed.
Fig. 6: Modelling results
ZF (m)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Y (
vol%
)
0
2
4
6
8
10
12
14
16
ZF (m)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Y (
vol%
)
0.0
0.1
0.2
0.3
0.4
0.5
0.6
C3H8
O2
CO2
(a) (b)
165
5. CONCLUSIONS
In the present study, the combustion of propane in the freeboard of a fluidized bed was studied by injecting propane
inside a high temperature fluidized bed through a downward facing sparger. The freeboard region was characterized
using pressure, solids flux, species volume fraction and temperature measurements during combustion. The
combustion in the fluidized bed and freeboard was confirmed by a decrease in oxygen and propane as well as a
temperature increase in the reactor. Solids flux and species volume fractions were measured simultaneously using a
non-isokinetic probe. The measurements showed a net upward flow of solids at the centre of the reactor and a net
downward flow near the wall. Furthermore, the upward solids flux was found to decrease exponentially with height
from the bed surface. The measured entrainment at the bed surface was found to be significantly higher than
predicted by the correlation of Wen et al. (1982): 17.6 kg/m2s compared to 8.18×10-4 kg/m2s. Pressure and solids
flux measurements were also used to locate the fluidized bed surface. Solids flux measurements were found to be
more precise than pressure to determine bed height. The location of the bed surface was taken as the inflection point
of the entrainment curve. The temperature profile corresponding to the measured solids flux and species volume
fractions was determined using a series of thermocouples positioned along the length of the reactor.
REFERENCES
Aguillón, J., Shakourzadeh, K. and P. Guigon (1995) Comparative study of non-isokinetic sampling probes for solids flux measurement in circulating fluidized beds. Powder Technol. 83, 79-84.
Botella, P., Lopez Nieto, J.M. and B. Solsona (2002). Selective oxidation of propene to acrolein on Mo-Te mixed oxides catalyst prepared from ammonium telluromolybdates. J. Mol. Catal. A-Chem. 184, 335-347.
Burcat, A. and A. Lifshitz (1971) Shock-tube investigattion of ignition in propane-oxygen-argon mixtures. In: Proceedings of the 13th International Symposium on Combustion, 745-755, Combustion Institute, Pittsburgh.
Chen, L.H. and C.Y. Wen (1982) Model of solid gas reaction phenomena in the fluidized bed freeboard. AIChE J. 28, 6, 1019-1027.
Dounit, S., Hemati, M. and D. Steinmetz (2001). Natural gas combustion in fluidized bed reactors between 600 and 850ºC: experimental study and modelling of the freeboard. Powder Technol. 120, 49-54.
Dounit, S., Hemati, M. and R. Andreux (2008). Modelling and experimental validation of a fluidized-bed reactor freeboard region: Application to natural gas combustion. Chem. Eng. J. 140, 457-465.
Gordon S. and B.J. McBride (1994). Computer program for calculation of complex chemical equilibrium compositions and applications. NASA RP-1311.
Hesketh, R.P. and J.F. Davidson (1991). Combustion of methane and propane in an incipiently fluidized bed. Combust. Flame. 85, 449-467.
Kunii, D. and O. Levenspiel (1991). Fluidization Engineering. 2nd edition. Boston, MA: Butterworth-Heinemann. Kim, K. and K. Soo Shin (2001). Shock tube and modeling study of the ignition of propane. Bull. Korean Chem.
Soc. 22, 3, 303-307. Large J.F., Martinie, Y. and M.A. Bergougnou (1977). Interpretative model for entrainment in a large gas-fluidized
bed. J. Powders Bulk Solids Technol. 1, 15-21.
166
Malleswara Rao, T.V. and G. Deo (2007). Ethane and propane oxidation over supported V2O5/TiO2 catalysts: analysis of kinetic parameters. Ind. Eng. Chem. Res. 46, 1, 70-79.
Marinov, N.M., Pitz, W.J., Westbrook, C.K., Vincitore, A.M., Castaldi, M.J., Senkan, S.M. and C.F. Melius (1998). Aromatic and polycyclic aromatic hydrocarbon formation in a laminar premixed n-Butane flame. Combust. Flame. 114, 1-2, 192-213.
Morooka S., Kawazuishi, K. and Y. Kato (1980). Holdup and flow pattern of solids particles in freeboard of gas-solid fluidized bed with fine particles. Powder Technol. 26, 75-82.
Patience, G.S. and M.J. Lorences (2006). VPO transient oxidation kinetics. Int. J. Chem. React. Eng, 4, A22. Penyazkov, O.G., Ragotner, K.A., Dean, A.J. and B. Varatharajan (2005). Autoignition of propane-air mixtures
behind reflected shock waves. In: Proceedings of the Combustion Institute, 1941-1947, Elsevier Inc. Reinhardt, B., Cordonnier, A. and P. Florent (1999). Use of an isokinetic sampling propane. Results in a cyclone.
Powder Technol. 101, 81-90. Rhodes, M.J., Laussmann, P., Villain, F. and D. Geldart (1988). Measurement of radial and axial solids flux
variations in the riser of a circulating fluidized bed, In: Circulating Fluidized Bed Technology II (Basu, P. and J.F. Large (Eds)), 155-164, Pergamon Press, Oxford.
Rhodes, M.J. and P. Laussmann (1992). A simple non-isokinetic sampling probe for dense suspensions. Powder Technol. 70, 141-151
Van Breugel, J.W., Stein, J.J.M. and R.J. de Vries (1970). Isokinetic sampling in a dense gas-solids stream. Proc. Instn. Mech. Engrs. 184, 18-23.
Van der Vaart, D.R. (1988). Freeboard ignition of premixed hydrocarbon gas in a fluidised bed. Fuel, 67, 1003-1007. Van der Meer, E.H., Thorpe, R.B. and J.F. Davidson (2000). Flow patterns in the square cross-section riser of a
circulating fluidized bed and the effect of riser exit design. Chem. Eng. Sci. 55, 4079-4099. Wen, C.Y. and L.H. Chen (1982). Fluidized bed freeboard phenomena: entrainment and elutriation. AIChE J. 28, 1,
117-128. Werther, J. (1999). Measurement techniques in fluidized beds. Powder Technol. 102, 15-36.
167
ANNEXE 5 – ARTICLE PRESENTÉ À LA CONFÉRENCE
FLUIDIZATION XIII (CORÉE DU SUD, 2010)
FLUIDIZED BED COMBUSTION OF C1-C4 N-ALKANES
Jean-Philippe Laviolette, Gregory S. Patience and Jamal Chaouki*
Department of Chemical Engineering, École Polytechnique de Montréal, P.O. Box 6079, Station
Centre-ville, Montréal, Canada H3C 3A7
*Corresponding author. Tel.: +1 514 340 4711 ext. 4469; fax: +1 514 340 4159.
E-mail address: [email protected] (J. Chaouki)
Abstract: The non-premixed combustion of C1 to C4 n-alkanes was investigated inside a
bubbling fluidized bed of inert sand particles at intermediate temperatures: 923 K
(650oC) ≤ TB ≤ 1123 K (850oC). Lower (T1) and upper (T2) critical transition bed temperatures
were measured for ethane, propane and n-butane as 923 K (650oC) and 1073 K (800oC),
respectively. The values for methane were significantly higher with T1 = 1023 K (750oC) and
T2 > 1123 K (850oC). Alkane conversion was accurately modeled with first-order kinetics and C2
to C4 n-alkanes combustion rates were characterized by a uniform Arrhenius expression:
� � �2.59 � 10�� �� �������� ��� . The reaction rate of methane was significantly slower with:
���� � �1.99 � 10��� ���� �������� ��� .
Keywords: combustion, fluidized bed, methane, propane, ethane and n-butane
1. Introduction
The combustion of methane, LPG and propane in fluidized beds of inert particles has been the
subject of several studies in premixed [1-10] and non-premixed modes [11-16]. Fluidized bed
temperature has been reported as a key parameter that determines in-bed fuel conversion:
combustion is initiated deeper inside the fluidized bed with increasing temperature. In premixed
combustion, increasing temperature has been shown to move the combustion front towards the
distributor - the combustion front moves first from the freeboard to the bubbles bursting at the
upper surface of the bed, then to bubbles igniting in the bed and finally to small bubbles where
168
ignition occurs practically at the level of the distributor and the process appears flameless [7, 9].
While some studies have suggested that combustion did not occur to any significant extent
inside the emulsion phase [9], methane and propane combustion have been observed in fixed
beds of sand particles at temperatures above 1023 K (750oC) and 973 K (700oC),
respectively [1, 17].
Two critical transition temperatures have been defined for fluidized bed combustion in inert
particles. Below the lower critical temperature (T1), combustion only occurs above the bed
surface. Between T1 and the upper critical temperature (T2), combustion begins to move inside
the bubbles within the bed. Finally, above T2, combustion takes place entirely within the bed and
close to the distributor. For methane combustion in inert sand particles, the values of T1 and T2
reported in the literature range between 823-1018 K (550-745oC) and 1133-1250 K (860-977oC),
respectively [18]. For propane, T2 was measured in a shallow bed (HB ≈ 0.1 m) of sand particles
as 1123 K (850oC) [2, 9]. Investigations on the fluidized bed combustion of other alkanes in inert
particles are scarce.
In-bed conversion for specific operating condition can also be estimated by reaction models that
combine gas/solids hydrodynamics and reaction kinetics. Several microkinetic and first-order
global kinetics schemes are available in the literature for gas-phase systems [5, 6, 19]. Gas-
phase microkinetic models have also been modified to account for the “quenching” effect of
solids particles on the free radicals [17, 20]. However, microkinetic models are very complex,
while global kinetic models are valid only for a specific range of operating conditions such that
both types of models may yield significant uncertainties. Furthermore, using global kinetics
schemes for different hydrocarbons from various sources and measured under different
experimental conditions may result in further errors in the modeling results.
It may be possible to express the first-order kinetics and conversion of C2 and higher alkanes in
terms of a single expression or correlation. Studies have reported similarities between the
combustion chemistry of n-alkanes: their combustion rate is controlled by the same main chain
branching path, which can be summarized by the following elementary reactions at intermediate
temperatures (~ 850 K (577oC) to 1200 K (927oC)) [21-24]:
� � � �! " � � �! (1)
169
#� � � � " # ��� � (2)
�� � �! " � � � �! (3)
In the above equations, RH is an alkane, R is an alkyl radical and M is a third body.
Saxena et al. [23] have demonstrated that the high temperature induction time of propane and
higher alkanes can be accurately estimated by employing rate parameters of elementary steps
(2) to (4).
� � �� � " �� � � � (4)
Consequently, the induction time of C2 and higher n-alkanes have been shown to exhibit similar
dependence on temperature, pressure and mixture composition such that it may be calculated
from a single correlation with an uniform activation energy [21-22]. Methane, on the other hand,
has been observed to yield significantly higher induction times, which can be explained by the
production of a methyl radical in equation (2), which recombines to form ethane, resulting in
chain termination. If the combustion chemistry and induction time correlations among C2 and
higher n-alkanes show such similarities, it seems reasonable to think that a parallel may also be
made between the global reaction kinetics and conversion.
In the present study, the non-premixed combustion of methane, ethane, propane and n-butane
was studied in a fluidized bed of sand particles at temperatures ranging from 923 K (650oC) to
1123 K (850oC). The flow of each hydrocarbon was adjusted to obtain a constant heat release
rate of 3.8 kW and critical bed temperatures were measured for each hydrocarbon. Furthermore,
a reaction model with first-order reaction kinetics was developed and compared to the
experimental results. The first-order kinetic parameters were measured for each n-alkane and
compared.
2. Experimental
The experiments were performed in a fluidized bed reactor with an inner diameter of 0.2 m and a
height of approximately 1.0 m. Inert sand particles (dp = 290 µm and Umf = 0.08 m/s) were used
as the bed material. The bed was fluidized with air, which was injected through the distributor,
and the hydrocarbon fuels (methane, ethane, propane and n-butane) were separately injected
through a downward-facing sparger whose the tip was located at a distance of 0.1 m above the
170
distributor. Prior to each experiment, the fluidized bed reactor was heated to the desired
temperature with a natural gas burner connected to the windbox.
The experiments were conducted at a superficial gas velocity of 0.4 m/s and at five bed
temperatures: 923 K (650oC), 973 K (700oC), 1023 K (750oC), 1073 K (800oC) and 1123 K
(850oC). The temperature was measured at 9 different axial positions with thermocouples. Gas
was sampled at six axial and five positions. The sampled gas was analyzed by a CO/CO2
analyzer and a micro-GC connected in series to measure the volume fraction of the following
species: H2, CH4, C2H4, C2H6, C3H8, n-C4H10, O2, N2, CO and CO2. At each axial position, a
cross-sectional average of species volume fraction was calculated from the measurements. The
mixed-cup volume fraction at the injector tip was adjusted for each hydrocarbon to obtain a
constant heat release rate of 3.8 kW. For methane, ethane, propane and n-butane, the mixed-
cup volume fraction was 5.1%, 2.5%, 1.9% and 1.3%, respectively.
3. Results and discussion
3.1 Critical transition fluidized bed temperatures
Lower (T1) and upper (T2) critical transition bed temperatures were measured for methane,
ethane, propane and n-butane. Figure 1a shows the axial profile of conversion for ethane,
propane and n-butane inside the fluidized bed reactor at a bed temperature of 923 K (650oC).
The corresponding axial profile of CO2 volume fraction is presented in Figure 1b. The bed height
was 0.55 m and it is shown on both figures as a dashed line. At 923 K (650oC), ethane, propane
and n-butane conversion was not significant inside the fluidized bed: conversion fluctuated
between 0-4%, 0-18% and 10-27% for ethane, propane and n-butane, respectively.
Furthermore, the CO2 volume fractions in the fluidized bed remained low at approximately 0.7%.
Combustion occurred mainly at the bed surface and inside the freeboard as conversion reached
92%, 93% and 96% at 0.2 m above the bed surface (Z = 0.75 m) for ethane, propane and
n-butane, respectively. This was accompanied by a significant increased in the CO2 volume
fraction, which reached approximately 5.5% for all alkanes. Complete conversion was obtained
at Z = 0.97 m. Since in-bed conversion was limited and most hydrocarbons burned inside the
freeboard, the lower critical temperature (T1) was defined as 923 K (650oC) for ethane, propane
and n-butane.
171
As the bed temperature was increased to 973 K (700oC), the combustion front moved upstream
inside the fluidized bed for ethane, propane and n-butane. Figures 1c and 1d show the axial
profile of conversion and CO2: combustion was initiated within 0.15 m (Z = 0.25 m) of the injector
tip where a conversion of 24%, 57% and 61% was measured for ethane, propane and n-butane,
respectively. Conversion increased steadily in the fluidized bed and partial in-bed conversions
of 60%, 79% and 84% were measured at 0.1 m below the bed surface (Z = 0.46 m) for ethane,
propane and n-butane. The unconverted hydrocarbons burned inside the feeboard region and
complete conversion was observed at 0.2 m above the bed surface (Z = 0.75 m). The measured
conversion in the fluidized bed and freeboard regions was accompanied by a corresponding
increase in the CO2 volume fraction, which reached approximately 3.8% at 0.1 m below the
fluidized bed surface (Z = 0.46 m) and 6.0% in the freeboard region at Z = 0.97 m.
Figures 1(e) and 1(f) show the axial profile of CO2 and conversion for the three n-alkanes as the
bed temperature was increased further to 1023 K (750oC). Near complete fuel conversion
occurred in the fluidized bed with ethane, propane and n-butane: conversion reached 95%, 95%
and 98%, respectively, only 0.36 m (Z = 0.46 m) downstream of the injector. At that position, the
CO2 volume fraction reached about 6.0%, which corresponded approximately to the theoretical
value at 100% combustion. Complete conversion was observed 0.2 m over the bed surface
(Z = 0.75 m).
At a bed temperature of 1073 K (800oC), the conversion of ethane, propane and n-butane,
conversion reached 99% within 0.25 m of the sparger tip (Z = 0.35 m): this temperature was
defined as the upper critical temperature (T2). This value is lower than the 1123 K (850oC)
previously reported for propane [2, 9]. However, this discrepancy may be explained by the fact
that conversion was not measured at a bed temperature of 1073 K (800oC) – experiments were
performed at bed temperatures of 1023 K (750oC) and 1123 K (850oC) [2, 9].
Methane fluidized bed combustion required significantly higher bed temperatures compared to
ethane, propane and n-butane. Figure 2 shows the axial profile of methane conversion at four
bed temperatures: 973 K (700oC), 1023 (750oC), 1073 (800oC) and 1123 K (850oC). At a bed
temperature of 973 K (700oC), no methane conversion was measured in the fluidized bed and
172
combustion occurred entirely in the freeboard region: methane conversion increased from 37%
at 0.06 m above the bed surface (Z = 0.61 m) to 99% at Z = 0.97 m, respectively.
Figure 1: Conversion and YCO2 axial profiles at 923 K and 973 K
Z (m)
0.0 0.2 0.4 0.6 0.8 1.0
Xi (
%)
0
20
40
60
80
100
C2H
6
C3H
8
C4H
10
TB = 923 K (650
oC)
Z (m)
0.0 0.2 0.4 0.6 0.8 1.0
YC
O2 (
vo
l%)
0
2
4
6
8
10
C2H
6
C3H
8
C4H
10
TB = 923 K (650
oC)
Z (m)
0.0 0.2 0.4 0.6 0.8 1.0
Xi (
%)
0
20
40
60
80
100
C2H
6
C3H
8
C4H
10
TB = 973 K (700
oC)
Z (m)
0.0 0.2 0.4 0.6 0.8 1.0
YC
O2 (
vo
l%)
0
2
4
6
8
10
C2H
6
C3H
8
C4H
10
TB = 973 K (700
oC)
Z (m)
0.0 0.2 0.4 0.6 0.8 1.0
Xi (
%)
0
20
40
60
80
100
C2H
6
C3H
8
C4H
10
TB = 1023 K
(750oC)
Z (m)
0.0 0.2 0.4 0.6 0.8 1.0
YC
O2 (
vo
l%)
0
2
4
6
8
10C
2H
6
C3H
8
C4H
10
TB = 1023 K
(750oC)
(a) (b)
(c) (d)
HB H
B
HB H
B
(e) (f)HB
HB
173
As the temperature was increased to 1023 K (750oC), figure 2 shows that the methane in-bed
conversion was small – the measured conversion fluctuated between 0% and 12% and the CO2
volume fraction reached about 2.0%. Combustion mainly occurred at the bed surface –
conversion increased to 64% at 0.06 m (Z = 0.61 m) above the bed surface. This was
accompanied by a significant increase in the volume fraction of CO2 in the freeboard. Complete
methane conversion was observed at Z = 0.97 m. A bed temperature of 1023 K (750oC) was
defined as the lower critical temperature (T1), which is similar to the observations of previous
studies [18].
At a higher bed temperature of 1073 K (800oC), significant in-bed conversion was observed:
conversion reached 23% at a distance of 0.15 m downstream of the sparger (Z = 0.25 m) and
increased to 40% at Z = 0.46 m. The unconverted methane burned in the freeboard and
complete conversion was obtained at Z = 0.97 m.
Figure 2: XCH4 axial profiles at various fluidized bed temperatures
The bed temperature was increased further to 1123 K (850oC) and metane conversion increased
to 83% at 0.1 m below the bed surface (Z = 0.46 m). The unconverted methane burned inside
the freeboard region and within 0.2 m of the bed surface. The upper critical temperature (T2) for
methane is higher than 1123 K (850oC) and this agrees with the reported values in the scientific
literature [18].
Z (m)
0.0 0.2 0.4 0.6 0.8 1.0
XC
H4 (
%)
0
20
40
60
80
100
973
1023
1073
1123
TB (K)
174
3.2 C1-C4 n-alkanes global reaction rates
A reaction model was developed and compared to the experimental measurements. The
fluidized bed gas-phase hydrodynamics and reaction kinetics were modeled as a plug flow
reactor and a first-order reaction, respectively. The hydrocarbon conversion was calculated from
equations (5) and (6):
$� � 100 %1 & ��'′()*+,�-*./ (5)
0 ′ � 1'23 ��45 ��� � 0�′ ��
45 ��� (6)
In equation (5), the parameter bi represents an induction delay: bi increases with induction time.
Furthermore, Zinj, k, A, v, EA and R corresponds to the height above the sparger tip (Zinj = 0 m
corresponds to Z = 0.1 m), the kinetic constant, the fluidized bed cross-sectional area, the
volume flow rate in the fluidized bed, the activation energy and the universal gas constant. The
use of this simple reaction model is consistent with the work of Lorences et al. [25], which have
shown that the gas-phase hydrodynamics in fluidized beds are very close to plug flow when
Ug ≤ 6 × Umf. Also, several fluidized bed combustion studies have reported that fuel-lean
combustion can be accurately represented as first order with respect to hydrocarbon
concentration [17, 26-27].
Figure 3a shows the in-bed conversion of ethane, propane and n-butane as a function of Zinj on
a natural logarithmic scale for two bed temperatures: 973 K (700oC) and 1023 K (750oC). A
linear relationship is observed between the ordinate [ln(1-Xi)] and Zinj, which indicates that the
model agrees well with the experimental results. The slope of the curve represents the first order
reaction rate (k’): the same reaction rate was measured for ethane, propane and n-butane. For
bed temperatures of 973 K (700oC) and 1023 K (750oC), k’ was measured as 3.8 m-1 and 8.2 m-
1, respectively. Therefore, the temperature dependence of the reaction rate was uniform for C2 to
C4 n-alkanes, which implies that the activation energy (EA) was also the same. However, the
parameter bi in equation (6) was not uniform among C2 to C4 n-alkanes. Table 1 lists the
different values of bi measured at the different bed temperatures and used in the fluidized bed
combustion model. Figure 3b compares the measured and predicted conversion of ethane,
propane and n-butane as a function of (ZI + b): the agreement between the model and the
experimental results is very good.
175
Figure 3: Comparison between the model and experiments at 973, 1023,1073 and 1123 K
Similar observations were made at higher bed temperatures. Figure 3c shows the in-bed
conversion of methane, ethane and propane as a function of Zinj on a natural logarithmic scale
for two bed temperatures: 1073 K (800oC) and 1123 K (850oC). The first-order global reaction
rate for ethane and propane was measured as 13.5 m-1 at 1073 K (800oC). N-butane conversion
reached 100% and could therefore not included in Figure 3c. For methane, however, the
reaction rate was significantly lower compared to C2-C4 n-alkanes: k’ rose from 1.2 m-1 to 4.2 m-1
as the bed temperatures was increased from 1073 K (800oC) to 1123 K (850oC). Figure 3d
compares the measured and predicted conversion of methane, ethane, propane and n-butane in
the fluidized bed: at these higher bed temperatures, the agreement is also very good.
Zinj
(m)
0.1 0.2 0.3 0.4
(1-X
i)
e-6
e-5
e-4
e-3
e-2
e-1
e0
C2H
6
C3H
8
C4H
10
Zinj
(m)
0.1 0.2 0.3 0.4
(1-X
i)
e-6
e-5
e-4
e-3
e-2
e-1
e0
CH4
C2H
6
C3H
8
k'1 = 3.8 m
-1
k'2 = 8.2 m
-1
k'3 = 13.5 m
-1
k'4 = 1.2 m
-1
k'5 = 4.2 m
-1
Zinj
+ bi (m)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Xi (
%)
0
20
40
60
80
100
C2H
6
C3H
8
C4H
10
Model
TB = 973 K
(700oC)
(a) (b)
Zinj
+ bi (m)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Xi (
%)
0
20
40
60
80
100
CH4
C2H
6
C3H
8
C4H
10
Model
TB = 1123 K
(850oC)
TB = 1073 K (800
oC) (d)(c)
TB = 1023 K
(750oC)
TB = 973 K (700
oC)
TB = 1023 K
(750oC)
TB = 1073 K (800
oC)
TB = 1123 K (850
oC)
176
It is observed in Table 1 that the magnitude of bi decreased with increasing alkane carbon
number at a constant bed temperature. For methane and ethane, bi decreased linearly with
temperature, which is consistent with induction time studies: induction time decreased with
increasing temperature. For ethane, bi decreased from 0.085 m at 973 K (700oC) to -0.045 m at
1073 K (800oC). A negative bi can be explained by a significant hydrocarbon conversion near the
injector tip at high bed temperatures. Studies have shown that the jet originating from an injector
enhances mixing, which can lead to significant conversion in that region when the mixture
induction time is sufficiently low [18]. For methane, bi decreased from 0.075 m at 1073 K (800oC)
to -0.05 m at 1123 K (850oC). For propane and n-butane, bi remained approximately constant at
-0.05 m and -0.1 m, respectively.
Table 1: Parameter bi at the different bed temperatures (TB)
n-Alkane bi (m) Range (K)
CH4 -0.0025 TB + 2.76 1073-1123
C2H6 -0.0013 TB + 1.35 973-1073
C3H8 -0.05 973-1073
n-C4H10 -0.1 973-1073
The global reaction rates can be expressed in the Arrhenius form:
0���′ � �1.99 � 10��� �������� ��� (7)
0���6′ � 0����′ � 0�����′ � �2.59 � 10�� �������� ��� (8)
The reaction rate constant and the activation energy for methane were calculated from the
reactions rates measured at bed temperatures of 1073 K (800oC) and 1123 K (850oC). On the
other hand, the parameters for ethane, propane and n-butane were calculated from the
reactions rates at 973 K (700oC) and 1023 K (750oC). Equation (8) overpredicts the global
reaction rate at 1073 K (800oC) and yields 16.4 m-1 compared to the measured value of 13.5 m-1.
This discrepancy can be explained by the fact that the in-bed conversions were very high at this
177
temperature (in the order of 93% to 100%). At these high conversions, a variation of 1.0% in the
measured value greatly affected the measured slope in Figure 3c. The precision in the
conversion measurements was estimated for the present study as ±5% absolute.
The fact that ethane, propane and n-butane have the same global kinetics is consistent with
previous studies, which have shown that the induction times of C2 and higher alkanes are of
similar magnitude and are governed by the same temperature dependence or activation
energy [21-22]. Methane is also known to be significantly less reactive than higher n-alkanes
[22,24]. Table 1 shows that the parameter bi increased with increasing number of carbons in the
n-alkane molecule, which suggests that the mixture reactivity increased (and induction time
decreased) with increasing fuel molecule size. This trend is consistent with previous studies [21-
22] except for ethane, which was reported by Burcat et al. [22] to show the highest reactivity
among C1 to C5 n-alkanes. This was also supported by Westbrook [24] with kinetic
considerations. This discrepancy between the present results and the scientific literature can not
be explained and the ± 5% (absolute) error in the measured conversion may be at cause.
4. Conclusions
The non-premixed combustion of methane, ethane, propane and n-butane was investigated in a
fluidized bed of inert sand particles. The bed temperature was varied between 923 K (650oC)
and 1123 K (850oC) while the superficial gas velocity was kept at 0.4 m/s. Critical transition bed
temperatures (T1 and T2) were measured for each fuel. For ethane, propane and n-butane,
combustion occurred mainly in the freeboard region at bed temperatures below: T1 = 923 K
(650oC). On the other hand, complete conversion occurred within 0.2 m of the injector at:
T2 = 1073 K (800oC). For methane, the measured values of T1 and T2 were significantly higher at
1023 K (750oC) and above 1123 K (850oC), respectively. A reaction model with first-order
kinetics was developed and compared to the experimental measurements. The reaction rate of
C2 to C4 n-alkanes could be characterized by a single kinetic expression and a uniform reaction
rate constant and activation energy. On the other hand, the methane reaction rate was
significantly slower with a higher activation energy.
References
[1] Hesketh RP, Davidson JF. Combust. Flame 1991;85:449-467.
178
[2] Van der Vaart DR. Combust. Flame 1988;71:35-39.
[3] Van der Vaart DR. Fuel 1988;67:1003-1007.
[4] Friedman J, Li H. Combust. Sci. Technol. 2005; 177; 2211-2241.
[5] Dounit S, Hemati M, Andreux R. Chem. Eng. J. 2008; 140; 457-465.
[6] Pre P, Hemati M, Marchand B. Chem. Eng. Sci 1998; 53 (16); 2871-2883.
[7] Zukowski W. Combust. Flame 2003; 134; 399-409.
[8] Baron J, Bulewicz EM, Zukowski W, Kandefer S, Pilawska M. Combust. Flame 2002; 128; 410-421.
[9] Dennis JS, Hayhurst AN, Mackley IG. 19th Symp. (Intl.) on Combust. 1982; 1205-1212.
[10] Srinivasan RA, Sriramulu S, Kulasekaran S, Agarwal PK. Fuel 1998; 77; 1033-1049.
[11] Mabrouk R, Chaouki J, Laviolette J-P, Patience GS. Proc. 9th Conference on Circulating Fluidized
Beds, Hamburg, Germany, 2008.
[12] Ross DP, Yan HM, Zhang DK. Combust. Flame 2001; 124; 156-164.
[13] Sotudeh-Gharebaagh R. PhD Thesis, Ecole Polytechnique, Montreal, Canada, 1998.
[14] Singh B, Rigby GR, Callcott TG In: Inst. of Fuel Symp. Series, No.1, London (1975), C5.1–C5.9.
[15] Stubington JF, Davidson JF. AIChE J. 1981; 27 (1); 59-65.
[16] Sotudeh-Gharebaagh R, Chaouki J. Energy & Fuels 2007; 21; 2230-2237.
[17] Sotudeh-Gharebagh R, Chaouki J. Can. J. Chem. Eng. 2003; 81; 1182-1191.
[18] Laviolette J-P, Sotudeh-Gharebagh R, Mabrouk R, Patience GS, Chaouki J. In A. Agarwal (ed.)
Handbook of Combustion (vol.5), Wiley 2010.
[19] Ross DP, Yan H-M, Zhang D-K. Fuel 2004; 83; 1979-1990.
[20] Jeng R-S, Altwicker ER, Morgan III MM, Wilk RD. Combust. Sci. Technol. 2001; 170; 119-149.
[21] Horning DC, Davidson DF, Hanson RK. Journal of Propulsion and Power 2002; 18 (2); 363.
[22] Burcat A, Scheller K, Lifshitz A. Combust. Flame 1971; 16; 29-33.
[23] Saxena P, Peters N, Williams FA. Combust. Flame 2007; 149; 79-90.
[24] Westbrook, CK. Proc. Combust. Inst. 2000; 28; 1563-1577.
[25] Lorences M, Laviolette J-P. Powder Technol. 2006; 168 (1); 1-9.
[26] Foka M, Chaouki J, Guy C, Klvana D. Chem. Eng. Sci. 1994; 49 (24A); 4269-4276.
[27] Chaouki J, Klvana D, Guy C. Korean J. Chem. Eng. 1999; 16 (4); 494-500.
179
ANNEXE 6 – ARTICLE PRÉSENTÉ À LA CONFÉRENCE
FLUIDIZATION XIII CONFERENCE (CORÉE DU SUD, 2010)
FIBRE-OPTIC PROBE FOR THE SIMULTANEOUS MEASUREMENT OF
GASEOUS SPECIES COMPOSITION AND SOLIDS VOLUME FRACTION
Jean-Philippe Laviolette, Gregory S. Patience and Jamal Chaouki*
Department of Chemical Engineering, École Polytechnique de Montréal, P.O. Box 6079, Station
Centre-ville, Montréal, Canada H3C 3A7
*Corresponding author. Tel.: +1 514 340 4711 ext. 4469; fax: +1 514 340 4159.
E-mail address: [email protected] (J. Chaouki)
Abstract
A novel infrared fibre-optic probe was developed to measure quantitatively and simultaneously
solids volume fraction (1-ε) and gaseous species composition (Yi) in a gas/solid fluidized bed.
The fibre-optic probe was used with a FT-IR spectrometer to perform real-time and in-situ
measurements of absorbance in the fluidized bed. The effect of (1-ε) and Yi on the absorbance
spectra were additive and could be independently calibrated. To calibrate the probe, fuel mole
fractions and (1-ε) were varied between 1.8 - 10.1 mol% and 0 - 0.45, respectively. A proof of
concept for a novel application in fluidized beds was completed: the fibre-optic probe was used
to measure the molar fraction of a tracer gas inside the emulsion and bubble phases during gas
tracer experiments.
Keywords: solids volume fraction, FT-IR spectroscopy, fibre-optic probe, gaseous species
composition, fluidization, bubble, emulsion
1. Introduction
Gas/solid fluidized bed processes are characterized through the measurement of gaseous
species composition and solid volume fraction. These two parameters are dependent
parameters that are coupled through the reaction kinetics and hydrodynamics. Solids volume
fraction affects chemical reactions through catalytic and/or inhibitive effects as well as through
180
the thermal balance. For example, the chemical products from fast reactions, such as oxidation,
are greatly affected by the gas/solid mixing at the tip of injectors. Furthermore, at the
hydrodynamic level, fluidized beds are characterized by the emulsion and bubble phases:
determining the species composition in these two regions is key to the process characterization.
Current measurement methods are incapable of measuring gas-phase chemical composition
and solids volume fraction simultaneously and in-situ in a gas/solid fluidized beds. Gas-phase
chemical composition is usually measured with sampling probes connected to analyzers.
Furthermore, solids volume fraction is generally measured with capacitance probes and fibre-
optic probes that record the forward- or back-scattering of visible light [1-4].
It may be possible to use a fibre-optic probe with infrared (IR) light to measure solids volume
fraction and gas-phase chemical composition simultaneously and in-situ. IR fibre-optic probes
used with IR spectroscopy have been previously used to measured chemical composition in
multiphase systems [5-14]. However, current IR spectroscopic applications do not measure
solids volume fraction and rely on the uniformity of the solid samples with time (between
measurements) since the solid phase has been shown to affect the IR absorbance spectra. The
solids volume fraction and particle size influence the interaction between the light and the solids,
which affect the path length of the IR beam and the effective sample size [15-16]. However,
since multiphase systems are generally characterized by solids volume fractions that are
heterogeneous in space and time, in-situ and real-time measurement of species concentrations
in the different regions of a multiphase system requires the simultaneous determination of solids
volume fraction. The measurement volume can be made independent of the solid properties by
inclining the emitting and receiving fibres to form a convergent scheme [17] and this principle
can be applied to an IR fibre-optic probe to measure species concentrations.
In multiphase systems, the movement of powders can make spectroscopy measurements and
interpretation more complex. Studies have shown that the movement of the particles produced
artefacts on the measured spectrum at certain wavenumbers that depended on the timescale of
the spectral scan compared to the rate of movement of the particles. The wavenumber at which
the artefact appeared increased with increasing modulation frequencies such that the effects of
181
moving particles could be completely eliminated at the wavenumbers of interest with a
sufficiently high modulation frequency [18-20].
In the present study, a IR fibre-optic probe was constructed to perform in-situ and real-time
measurements in a gas/solid flow of methane/nitrogen and FCC particles (dp = 83 µm). The
fibre-optic probe was connected to a near- and mid-IR Fourier transform transmission
spectrometer (6000-1000 cm-1). Absorbance spectra were recorded at a frequency of 4.5 Hz for
a period of 75 seconds and both Yi and (1-ε) were evaluated from each spectrum. Methane
molar fractions in nitrogen were measured over a range of 0-10.1 vol%. Furthermore, the solids
volume fraction was varied between 0-0.45.
Figure 1: Fluidized bed apparatus
2. Experimental
Experiments were conducted with a mid-IR fibre-optic probe connected to FT-IR (Varian
Excalibur series 3100). The probe was composed of two parallel fluoride glass fibre-optics (an
emitting and a receiving fibre-optic) with a numerical aperture of 0.2 and a core diameter of
600 µm. The fibre-optic probe was connected to the FT-IR with a Harrick Fibremate™. A planar
gold-coated mirror was positioned perpendicularly to the probe tip at a distance of 5 mm. The
mirror reflected the emitted IR beam to the receiving fibre-optic, which resulted in a
CH4/N2
N2
IR fibre-optic probe
FT
-IR Gold-coated planar mirror
Bubble FCC
Porous plate
182
measurement volume between the probe tip and the mirror. Experiments were conducted by first
measuring a background spectrum - a CH4/N2 mixture was injected inside the measurement and
a background spectrum was obtained from the co-addition of 20 spectra. Then, the
measurement volume was flushed with air and the funnel was filled with sand particles to initiate
a flow of solids. Absorbance spectra were measured at a frequency of 4.5 Hz (temporal
resolution of 0.22 s) and the methane molar fraction was measured simultaneously to the solids
volume fraction.
Gas tracer experiments were conducted in a small scale fluidized bed reactor (I.D. = 5 cm)
illustrated in Figure 1 and the fibre-optic probe was used to measure the tracer molar fraction
inside the bubble and emulsion phases. The reactor was filled with FCC particles (dp = 83 µm
and 14% fines (dp ≤ 44 µm)) and was fluidized with nitrogen at a superficial gas velocity of
2.6 mm/s (Umf = 2.5 mm/s and Umb = 2.7 mm/s). The expanded bed height was 12.5 cm and the
fibre-optic probe was inserted in the bed at a height of 7 cm. A mirror was positioned
perpendicularly at the probe tip and gas bubbles were produced in the fibre-optic probe
measurement volume by injecting a mixture containing 10.1% CH4 + 89.9% N2 through a
downward facing sparger. The CH4/N2 mixture was injected at a flow rate of 10 mL/s by
manually opening a toggle valve for roughly 0.5 s at an interval of approximately 11 seconds.
3. Results and discussion
Both solid particles and gaseous species contribute to the absorbance spectrum during
transmission spectroscopy in a gas/solid sample: the total absorbance at wavelength λ has a
contribution from the solids volume fraction and from the gaseous species composition. Solid
particles reduce the incident beam intensity by absorption, reflection and diffusion. Solids can
also influence the path length and sample volume such that the absorbance due to the chemical
species is a function of solids volume fraction:
( ) ( )[ ]ελλελ −+= − 1,,1, fnAAAiYTotal (1)
In the specific case of transmission spectroscopy where the effect of solids fraction on path
length and sample volume can be neglected, equation (1) can be simplified to:
183
( ) λλελ ,,1, iYTotal AAA += − (2)
Figure 2: IR fibre-optic probe calibration
Equation (2) is valid when the reflected or diffused beam intensity measured is negligible
compared to the transmitted light intensity. In equation (2), the effects of solids volume fraction
and gaseous species composition on the absorbance spectrum are independent. Therefore, the
effect of (1-ε) and Yi can be calibrated independently. In the present experiments, the IR light
intensity from diffused reflectance was negligible such that equation (2) was valid. This was
verified experimentally be removing the mirror facing the probe tip.
A calibration was first performed with the fibre-optic probe to evaluate the effect of methane
molar fraction and solids volume fraction on the absorbance spectrum. The effect of methane
molar fraction (YCH4) was calibrated with four mixtures containing 0, 0.1, 3.25 and 10.1 vol% of
methane in nitrogen. The average absorbance of the methane peak (ACH4) was measured in the
region of 3018.96-3016.86 cm-1 and it is shown as a function of methane molar fraction in
Figure 2(a).
ACH4
0.0 0.1 0.2 0.3 0.4 0.5 0.6
YC
H4
(vo
l%)
0
2
4
6
8
10
12
ABaseline
0.0 0.2 0.4 0.6 0.8 1.0
(1-ε
)
0.00
0.02
0.04
0.06
0.08
0.10
(a) (b)
R2 = 0.9999 R
2 = 0.9814
( ) BA09579.01 =− ε23.25
8.504.0
CH4
CH4CH4
A
AY
+
+=
184
Figure 3: Time-history of (1-ε) and YCH4 at the fibre-optic probe tip
The effect of solids volume fraction was calibrated by producing a flow of solids in the fibre-optic
measurement volume with a funnel. The solids flow rate and solids volume fraction were varied
by modifying the diameter of the funnel throat. Four solids volume fractions were used (0, 0.015,
0.038 and 0.097) and the average baseline absorbance (A(1-ε)) was measured in the regions of
2997.73-2992.04, 3036.70-3030.72 and 3045.87-3040.39 cm-1 during 75 seconds at a frequency
of 4.5 Hz. The average baseline absorbance is shown in Figure 2(b) as a function of solids
volume fraction and the relationship is close to linear. Cutolo et al. [21] previously reported a
linear relationship between absorbance and solids volume fractions. In the present case, solids
volume fraction was measured by a conventional back-scattering fibre-optic probe with a
different measurement volume than the IR fibre-optic probe. Since the flow of particles was
heterogeneous, this may explain the observed deviation from the linear relationship.
Time (s)
0 20 40 60
(1-ε
)
0.0
0.1
0.2
0.3
0.4
0.5
YC
H4 (
%)
-10
-5
0
5
10
15
20
(1-ε)
YCH4
185
Figure 4: Measurement of YCH4 in the fluidized bed
Wavenumber (cm-1
)
28002900300031003200
A
-0.4
0.0
0.4
0.8
1.2
1.6
Wavenumber (cm-1
)
28002900300031003200
A
-0.4
0.0
0.4
0.8
1.2
1.6
Wavenumber (cm-1
)
28002900300031003200
A
-0.4
0.0
0.4
0.8
1.2
1.6(a)
(b)
(c)
CH4 peak
186
Gas tracer experiments were conducted in the fluidized bed of Figure 1 and the fibre-optic probe
was used with the spectroscopic method to measure the molar fraction of a CH4 tracer inside the
bubble and emulsion phases. The bed of FCC particles was incipiently fluidized with nitrogen
and bubbles were produced at the tip of the fibre-optic probe by the injection of a
nitrogen/methane mixture (10.1% CH4) through a sparger. Figure 3 shows the history of solids
volume fraction and methane molar fraction measured by the fibre-optic probe during a typical
experiment: the solids volume fraction was 0.45 in the emulsion phase and decreased
significantly to 0.05-0 when gas bubbles of CH4/N2 were injected at intervals of approximately 11
seconds. The measured methane volume fraction in the emulsion phase fluctuated significantly
(5% to -5%) due to the low intensity of the transmitted IR beam and the resulting low S/N.
Figure 4(a) shows a typical absorbance spectrum measured in the emulsion phase: the low
intensity of reflected signal resulted in a high absorbance throughout the spectrum. However,
the time-average methane molar fraction was measured accurately as 0.0% (100% nitrogen).
Figure 4(b) shows an absorbance spectrum measured inside a bubble: a methane peak was
observed at 3017.63 cm-1. However, the injection of bubbles produced gas/solids movement,
which caused noise in most recorded absorbance spectra: 5 spectra out of the 27 measured in
the bubble phase were sufficiently clear to make a measurement of YCH4. Figure 4(c) shows a
typical absorbance spectrum where noise was observed and the methane peak was
indistinguishable. This noise could be eliminated in the wavelengths of interest by using a FT-IR
with a higher modulation frequency.
The average peak height measured in the bubble phase corresponded to a methane molar
fraction of 5% compared to the injected 10.1% CH4 in nitrogen. This discrepancy may be due to
mixing at the injector tip between the gas injected through the sparger and the fluidizing gas.
Furthermore, the small number of clear spectra that were measured in the bubble phase could
also explain the low value of YCH4 measured since time-average measurements were more
accurate than instantaneous measurements. During the present study, the measured
absorbance spectra had to be exported and analyzed manually, which greatly reduced the
number of spectra that could be considered to measure YCH4 in the bubble phase. However,
these results clearly show that the developed spectroscopic method with a fibre-optic probe can
measure the gas composition in the bubble phase.
187
3.1 Limitation of the technique and future work
The fibre-optic probe used in this study is limited to ambient temperature applications: fluoride
glass fibre-optics used can be exposed to temperatures below 150oC. However, fibre-optic
probes have been developed with plastic fibre-optics for temperatures up to 1000oC [22].
Therefore, it is reasonable to assume that the measurement technique could be used to perform
in-situ measurements in high temperature multiphase systems with an adequately designed
fibre-optic probe.
This measurement technique is also limited by the modulation frequency of the spectrometer:
the modulation frequency needs to be sufficiently high to avoid noise in the absorbance
spectrum due to movement in the gas/solid system. The FT-IR used in the present study was
purchased in 2001 and had a maximum modulation frequency of 80 kHz. New models are
currently available with significantly higher modulation frequencies.
Another limitation of the technique is the IR beam intensity transmitted in the fibre-optic probe,
which needs to be as high as possible in order to maximize the signal-to-noise ratio. The IR
beam intensity can be increased by using a fibre-optic bundle, using a more sensitive detector,
using a more powerful IR source, optimizing the signal transmission at the probe tip and at the
FT-IR/fibre-optic probe interface (Harrick Fibremate™). More recent models of FT-IR also
propose higher IR source intensity compared to the equipment used in the present study.
4. Conclusions
A novel IR fibre optic probe was used to measure quantitatively and simultaneously solids
volume fraction (1-ε) and gaseous species composition (Yi) in a gas/solid fluidized bed. The to a
fibre-optic probe was connected to a FT-IR to perform real-time and in-situ measurements of
absorbance. The effect of solids volume fraction and gaseous chemical composition on the
absorbance spectra were additive and could be independently calibrated. Experiments were
conducted with methane/nitrogen and propane/nitrogen mixtures and FCC particles. Fuel mole
fraction and solids volume fraction were varied between 1.8 - 10.1 mol% and 0 - 0.45,
respectively. The fibre-optic probe was used in a fluidized bed to measure the molar fraction of a
188
gas tracer inside the emulsion and bubble phases during gas tracer experiments. The
measurement underestimated the methane volume fraction injected.
More measurements need to be performed in gas/solid systems at ambient and high
temperatures to fully demonstrate the possibilities of this measurement method. High modulation
frequencies from newer models of FT-IR should help resolve the problem of noise in the
absorbance spectra caused by gas/solid movement. In theory, this method could also be used
for liquid/solid, gas/liquid, liquid/liquid and gas/liquid/solid systems. However, work has to be
performed in order to identify the limitations in these specific systems. It seems also that the
accuracy of instantaneous fibre-optic probe measurements could be significantly improved by
increasing the IR beam intensity. This could be achieved by using a fibre-optic bundle, using a
more sensitive detector, using a more powerful IR source, optimizing the signal transmission at
the probe tip and at the FT-IR/fibre-optic probe interface (Harrick Fibremate™). New FT-IR
models also propose significantly higher IR signal intensity compared to the equipment used in
the present study.
NOTATION
A Absorbance
dp Average particle size (µm)
f Modulation frequency (kHz)
t Distance between probe tip and mirror (mm)
Yi Molar fraction of specie i (vol%)
(1−ε) Solids volume fraction
λ Wavelength (µm)
References
[1]Yates JG, Simons SRJ. Experimental methods in fluidization research. International Journal of
Multiphase Flow. 1994; 20: 297-330.
189
[2]Ishida M, Tanaka H. Optical probe to detect both bubbles and suspended particles in a three-
phase fluidized bed. Journal of Chemical Engineering of Japan. 1982; 15 (5): 389-391.
[3]Hu T, Yu B, Wang Y. Holdups and models of three phase fluidized beds. In: Ostergaard K,
Sorensen A. Fluidization V. New York: Engineering Foundation, 1986: 353.
[4]Esmaeili B, Chaouki J, Dubois C. An evaluation of the solid hold-up distribution in a fluidized
bed of nanoparticles using radioactive densitometry and fibre optics. Canadian Journal of
Chemical Engineering. 2008; 86: 543-552.
[5]Cogdill RP, Anderson CA, Delgado M, Chisholm R, Bolton R, Herkert T, Afnan AM, Drennen
III JK. Process analytical technology case study part I: feasibility studies for quantitative near-
infrared method development. AAPS PharmSciTech. 2005; 6 (2): E273-E283.
[6]Andersson M, Josefson M, Langkilde FW, Wahlund K-G. Monitoring of a film coating process
for tablets using near infrared reflectance spectrometry. Journal of Pharmaceutical and
Biomedical Analysis. 1999; 20: 27-37.
[7]Kirsch JD, Drennen JK. Determination of film-coated tablet parameters by near-infrared
spectroscopy. Journal of Pharmaceutical and Biomedical Analysis. 1995; 13: 1273-1281.
[8]Kirsch JD, Drennen JK. Near-infrared spectroscopic monitoring of the film coating process.
Pharmaceutical Research. 1996; 13 (2): 234-237.
[9]Abrahamsson C, Johansson J, Andersson-Engels S, Svanberg S, Folestad S. Time-resolved
NIR spectroscopy for quantitative analysis of intact pharmaceutical tablets. Analytical Chemistry.
2005; 77 (4): 1055-1059.
[10]Hailey PA, Doherty P, Tapsell P, Oliver T, Aldridge PK. Automated system for the on-line
monitoring or powder blending processes using near-infrared spectroscopy part I. system
development and control. Journal of Pharmaceutical and Biomedical Analysis. 1996; 14
(5): 551-559.
[11]Sonja Sekulic S, Wakeman J, Doherty P, Hailey PA. Automated system for the on-line
monitoring of powder blending processes using near-infrared spectroscopy: Part II. Qualitative
approaches to blend evaluation Journal of Pharmaceutical and Biomedical Analysis. 1998;
17: 1285-1309.
[12]Sonja Sekulic S, Ward HW, Brannegan DR, Stanley ED, Evans CL, Sciavolino ST, Hailey
PA, Aldridge PK. On-line monitoring of powder blend homogeneity by near-infrared
spectroscopy. Analytical Chemistry. 1996; 68: 509-513.
190
[13]Frake P, Greenhalgh D, Grierson SM, Hempenstall JM, Rudd DR. Process control and end-
point determination of a fluid bed granulation by application of near infra-red spectroscopy.
International Journal of Pharmaceutics. 1997; 151 (1): 75-80.
[14]Bellamy LJ, Nordon A, Littlejohn D. Real-time monitoring of powder mixing in a convective
blender using non-invasive reflectance NIR spectrometry. The Analyst. 2008; 133: 58-64.
[15]Armaroli T, Bécue T, Gautier S. Diffuse reflection infrared spectroscopy (drifts) : application
to the in situ analysis of catalysts. Oil & Gas Science and Technology. 2004 ; 59 (2): 215-237.
[16]Rantanen J, Räsänen E, Tenhunen J, Känsäkoski M, Mannermaa J-P, Yliruusi J. In-line
moisture measurement during granulation with a four-wavelength near infrared sensor: an
evaluation of particle size and binder effects. European Journal of Pharmaceutics and
Biopharmaceutics. 2000a; 50 (2): 271-276.
[17]Cui H, Mostoufi N, Chaouki J. Comparison of measurement techniques of local particle
concentration for gas-solid fluidization. In: Kwauk M, Li J, Yang W-C. Fluidization X. New York:
Engineering Foundation, 2001: 779-786.
[18]De Paepe ATG, Dyke JM, Hendra PJ, Langkilde FW. Rotating samples in FT-RAMAN
spectrometers. Spectrochimica Acta Part A. 1997; 53: 2261-2266.
[19]Berntsson O, Danielsson L-G, Folestad S. Characterization of diffuse reflectance fiber probe
sampling on moving solids using a Fourier transform near-infrared spectrometer. Analytical
Chimica Acta. 2001; 431: 125-131.
[20]Andersson M, Svensson O, Folestad S, Josefson M, Wahlund K-G. NIR spectroscopy on
moving solids using a scanning grating spectrometer – impact on multivariate process analysis.
Chemometrics and Intelligent Laboratory Systems. 2005; 75: 1-11.
[21]Cutolo A, Rendina I, Arena U, Marzocchella A, Massimilla L. Optoelectronic technique for the
characterization of high concentration gas-solid suspension. Applied Optics 1990, 29 (9),
pp. 1317-1322.
[22]Cui H, Sauriol P, Chaouki J. High temperature fluidized bed reactor: measurements,
hydrodynamics and simulation. Chemical Engineering Science. 2003; 58: 1071-1077.
191
ANNEXE 7 – CHAPITRE 9 DU « WILEY HANDBOOK OF
COMBUSTION » (VOLUME 5) (2010)
FLUIDIZED BED COMBUSTION OF NATURAL GAS AND OTHER HYDROCARBONS
Jean-Philippe Laviolette1, Rahmat Sotudeh-Gharebagh2, Rachid Mabrouk3, Gregory S. Patience1 and Jamal Chaouki1*
1Department of Chemical Engineering, Ecole Polytechnique, C.P. 6079, succ. Centre-Ville,
Montréal, H3C 3A7, Canada 2School of Chemical Engineering, College of Engineering, University of Tehran, P.O. Box
11155-4563, Tehran, Iran 3Haldor Topsøe, Nymøllevej 55, 2800 Lyngby, Denmark
* Corresponding author ([email protected])
1 INTRODUCTION
This chapter reviews key features of methane combustion in the presence of solid catalytic or non-catalytic particles in fluidized bed reactors. Catalytic combustion has been defined as a flameless process that proceeds at lower temperature compared to homogeneous combustion. Catalytic combustion also results in lower pollutant (NOX and CO) emissions and is not constrained by the flammability limits [1]. In the present work, the definition of catalytic combustion is broadened to include combustion with non-catalytic (inert) particles since it has been shown to satisfy many of the above criteria and since it is relevant to many applications.
Fluidized bed combustion is widely used in the chemical, petrochemical and energy industries and is central to many emerging technologies. Due to increasingly severe environmental regulations, increasing oil prices and the considerable natural gas world reserves, new technologies are currently under development such as:
• Co-firing of methane with different low calorific fuels (including solid) to increase
combustion efficiency and heat value • Methane combustion with solid catalytic and non-catalytic particles to minimize CO and
NOX in power and heat generation applications • Syngas production from the catalytic partial oxidation of methane • Chemical-looping combustion with a metal oxide oxygen carrier to eliminate NOX and
the cost of CO2 separation [2] • Biomass gasification and methane/biomass combustion to replace oil-based fuels
192
A thorough knowledge of fluidized bed combustion is essential for process control, optimization and minimization of pollutant emissions. In essence, fluidized bed combustion is governed by the reaction kinetics and the gas/solid hydrodynamics.
1.1 Heterogeneous and homogeneous kinetics
In the first section of this chapter, the intrinsic kinetics of combustion in gas/solid fixed beds are discussed. The intrinsic kinetics (without any limitations from the hydrodynamics) are characterized by simultaneous homogeneous and heterogeneous steps. Fixed bed reactors are generally used to study kinetics since the gas-phase hydrodynamics are considered to be ideal - plug flow - and mass transfer limitations are minimized when certain criteria regarding the ratio of particle diameter to tube diameter and to bed length are respected [3]. The kinetics are determined by the reaction conditions and the composition of the bed material. Three types of particles are considered in this chapter: combustion catalysts, inert particles (sand and alumina) and metal oxides. Several homogeneous and heterogeneous reaction networks have been published in the literature and range from simple single-step models to microkinetic models involving hundreds of elementary reactions. Several kinetic models are reviewed.
1.2 Mixing in gas/solid fluidized beds
Mixing is a critical phenomenon for fast and highly exothermic reactions, such as combustion, since it has a dominant influence on the prevailing mechanism and rate determining step. Although excellent gas-solid contact efficiency can be achieved in fixed bed reactors, the rise in bed temperature usually leads to the formation of hot spots, catalyst deactivation and even to conditions of homogeneous gas-phase combustion. Fluidized bed reactors are more attractive for highly exothermic reactions because of their superior heat management ability, proper mixing and excellent temperature control.
The schematic of a bubbling/turbulent fluidized bed reactor is shown in Figure 1.1. During operation, air is injected inside the windbox and then passes through the distributor (porous or perforated plate) to fluidize the bed of solid particles. The fuel can be premixed with air (premixed mode) or it can be separately injected inside the fluidized bed through a sparger (non-premixed mode). The gas/solid hydrodynamics depends strongly on the superficial gas velocity (Ug).
The bed of particles is in a fixed state (fixed bed) when Ug is below the minimum fluidization velocity (Ug < Umf) and the gas-phase hydrodynamics is close to plug flow. When Ug is increased and equals Umf, (Ug = Umf) the upward drag forces exerted by the gas on the particles becomes sufficiently strong to suspend the solids. At this state, the bed of particles is said to be fluidized since it exhibits fluidic behaviour. As Ug is increased further (Ug > Umf), bubbles of gas form at the distributor and rise inside the fluidized bed at several times the gas superficial velocity – this is the bubbling regime. Bubbles and the motion of solids result in gas by-passing and backmixing that produce a divergence from plug flow that amplifies with increasing superficial gas velocity
193
as the number and size of bubbles grow [4, 5, 6]. Particles disengage from the gas above the bed surface in the freeboard and disengagement regions. The cyclone collects the remaining solid particles in the off-gas and returns them to the fluidized bed through the dipleg.
At higher gas flow rates (Ug >> Umf), in the turbulent fluidization regime, solids entrainment becomes significant, the upper bed surface becomes diffuse and the fluidized bed is characterized by a turbulent motion of clusters of solids and voids of various shapes and sizes. With a further increase in gas flow rate (Ug >>> Umf), the solids are entrained out of the bed with the gas – this is lean-phase fluidization. Gas backmixing is significantly reduced in the lean fluidization regime. However, the void fraction increases considerably. A mode detailed review of fluidized beds can be found in Kunii et al. [6].
The main advantages of gas-solids fluidized bed reactors over fixed bed reactors are:
• Lower pressure drop (the pressure drop across the fluidized bed is constant and equals the
weight of the bed independently of the gas velocity) • Increased throughput • Ability to inject the reactants separately • Ability to add or remove particles continuously or intermittently without shutting down
the process • Improved heat transfer between the bed and the wall and immersed objects • Reduced axial and radial temperature gradients, minimizing the risk of hot spots • Increased catalyst effectiveness factors because of smaller particle sizes
However, fluidized bed reactors have some important disadvantages compared to fixed beds:
• Scale-up uncertainties due to more complex gas-solid hydrodynamics • Particle attrition • Wear on interior surfaces due to particle impingement • Particle entrainment requiring gas-solid separation equipment
Fluidized beds are generally operated in a non-premixed mode to avoid combustion in the regions upstream (windbox and piping) and to allow operation in the flammable region. Efficient gas/solid mixing is therefore critical to achieve high reactant conversion. In the second section of this chapter, the effects of the following parameters on gas/solid mixing are reviewed: superficial gas velocity, sparger flow pattern, sparger design, particle size, solid fraction and baffles.
1.3 Fluidized bed combustion
Premixed and non-premixed methane fluidized bed combustion with inert and catalytic particles has been the subject of many investigations [7, 8, 9, 10, 11, 12, 13, 14, 15]. High-combustion
194
efficiency can be achieved at a remarkably low temperature (Tb ≤ 1173 K) compared to conventional devices (T > 1473 K). The fluidized-bed reactors can be operated isothermally [16] and are capable of meeting all environmental requirements through the judicious selection of operating conditions and bed material [17]. In the third section of this chapter, the factors that influence methane conversion and pollutant emissions (CO and NOX) in fluidized bed combustion are reviewed.
1.4 Fluidized bed combustion modelling
The optimization and scale-up of fluidized bed combustion processes necessitates the development of reaction models that include the effects of both kinetics (kinetics submodel) and hydrodynamics (hydrodynamics submodel). Various models have been proposed to describe the performance of fluidized beds and these models differ greatly in complexity. It is important to adopt a model with the optimum degree of sophistication with sufficient complexity to capture the mechanistic elements affecting the main aspects of interest. However, models should not be so complex such that they contain features whose inclusion is unjustified. In the fourth section of the present chapter, several fluidized bed combustion models are reviewed.
2 HETEROGENEOUS AND HOMOGENEOUS KINETICS
In gas/solid systems, the reaction kinetics is characterized by simultaneous homogeneous and heterogeneous steps.
2.1 Homogeneous kinetics
Homogeneous kinetics can be described by a set of global kinetics or by elementary reactions (microkinetics).
2.1.1 Global homogeneous combustion kinetics
Single-step models [18], two-step models [18, 19, 20] and more complex models [18] are available from the scientific literature. The main advantage of global models is their simplicity. The addition of intermediate species such as CO and others along with further refinement of the reaction mechanism into several steps can make the predicted product composition more detailed at the expense of increased complexity and uncertainty due to more estimated parameters. The kinetics of Dryer et al. [19] is described by the following two-step reaction:
OHCOOCH CHr224 2
2
34 +→+ (2.1)
222
1COOCO COr→+ (2.2)
195
8.07.0
2444 OCHCHCH CCkr =
−−=
973
11000198exp235
4 TRkCH (2.3)
5.025.0
22 COOCOCOCO CCCkr =
−−=
973
11000171exp000371
TRkCO (2.4)
2.1.2 Microkinetic Combustion Models
Microkinetic models are constructed from a series of tens and sometimes hundreds of elementary reactions [21] and include many chemical species and free radicals. For example, the detailed Gas Research Institute mechanism (GRI-Mech release 3.0) for homogeneous combustion of methane consists of 325 elementary reaction steps and 53 species. Due to their complexity, microkinetic models allow more insight into reaction phenomena as the formation of many chemical products and intermediates can be tracked. In general, global models are unable to predict the formation of CO, NOX and other intermediates [22]. Furthermore, the microkinetic models have been shown to successfully reproduce several combustion phenomena such as temperature regions, cool flames, induction time, etc. On the downside, they require several years of development, can hardly be used to describe multiphase systems with complex hydrodynamics (for the case of fluidized beds, for example) and involve the estimation of hundreds of parameters, which results in significant uncertainties concerning the results.
Microkinetic models are available from several research group websites [23, 24]. The GRI mechanism has been developed specifically for methane combustion. It is a product of significant computational and experimental work. The validity of the GRI approach has been demonstrated for a wide range of conditions [25].
2.1.2 Reduced GRI Mechanism
Microkinetic models are unsuitable for detailed reaction engineering modeling because of their complexity. Simple models are more readily modified to account for the effect of solids – catalytic and non-catalytic – compared to the detailed GRI mechanism. However, microkinetic models may also be simplified by eliminating minor reactions while conserving the fast reactions such that the overall behaviour is conserved.
Not all the reactions contribute equally to the combustion process at a specific set of operating conditions. Some of them may contribute significantly, others marginally and the rest do not contribute at all. Some species reach almost stationary mole fractions via very rapid elementary steps, while for others, this change happens very slowly compared to the time scale of interest. Moreover, the components with a small net rate may cause severe problems when solving large systems of differential equations. These considerations have been previously used to lower the number of species and elementary reactions in the GRI model. Tomlin et al. [26],
196
Warnatz et al. [27], Frenklach [28], Warnatz [29] have reviewed the main mathematical tools for the construction, investigation and reduction of complex reaction mechanisms.
To develop a reduced reaction network, the forward and reverse reactions rates that are small compared to the overall reaction rate are eliminated. This can be achieved by collecting rate information during the simulations with the complete mechanism and by setting a threshold value δ [26]. The non-contributing reactions can be identified by evaluating the following criterion for all grid points in the simulation:
δ≤maxr
ri (2.5)
Chang et al. [30] and Sotudeh-Gharebagh et al. [31] used equation (2.5) to reduce the GRI mechanism release 1.2 and 2.11 with δ = 0.02 and δ = 0.005, respectively. The mechanism with δ = 0.02 contained a set of 104 reactions and 22 species. Using a smaller δ value of 0.005 produced a mechanism of 37 reactions and 17 species. This shows that the reduction technique is sensitive to the threshold value. Applying a uniform threshold value for different combustion conditions may either result in an oversimplified reaction scheme or the presence of non-contributing reactions. Therefore, a suitable δ threshold value should be chosen based on the combustion conditions.
The reduced mechanism of Sotudeh-Gharebagh et al. [31] is presented in Table 2.1. It was validated with methane-air combustion experiments performed in a flow reactor at the following conditions: 823 K ≤ T ≤ 1250 K and 0.5 ≤ φ ≤ 1. The C2 path was removed from the combustion chemistry. It is generally recognized in the scientific literature that the C2 path is a sequence of less well-understood steps and becomes important for sufficiently fuel-rich mixtures [32, 33].
2.2 Combined homogeneous and heterogeneous kinetics
Solid particles can influence the gas-phase kinetics through a heterogeneous catalytic and/or inhibitive effect. Catalysts can increase or decrease the rate of specific elementary reactions and promote the formation of specific products and intermediates (including free radicals). Catalytic effects can also occur through intermediate species interaction in the heterogeneous and/or homogeneous phases. On the other hand, inert particles affect the intrinsic kinetics by inhibiting reactions through radical recombination at their surface [10].
Solid particles also influence combustion phenomena by modifying the thermal balance of the system. Solid particles have a significantly higher heat capacity compared to gas. The thermal balance of the system plays an important role in combustion [32]. For example, a higher heat capacity can quench the thermal run-away preceding the autoignition of a fuel/oxygen mixture.
197
2.2.1 Inert Particles
Inert particles are extensively used for power generation since they are inexpensive, widely available and resistant to high temperatures. Combustion with inert particles is also relevant to other applications such as biomass combustion and gasification. Inert particles can considerably alter the combustion process due to heterogeneous reactions. Inert particles can inhibit combustion by quenching free radicals and the extent of inhibition varies with temperature [34].
Sotudeh-Gharebagh et al. [31] performed premixed combustion experiments in a fixed bed reactor of sand particles. Figure 2.1 shows methane conversion at the reactor outlet as a function of reactor temperature for the empty reactor and the reactor filled with sand particles of 523 µm. The mean gas residence time was 3 s for a feed gas containing 2 mol% of methane in air. Based on Figure 2.1, it can be concluded that:
(1) T < 1020 K: The methane conversion was higher in the presence of sand particles - sand
accelerated slightly the reaction.
(2) 1020 K < T < 1150 K: In this moderately high temperature range, the methane conversion was significantly lower in the presence of sand particles as they inhibited the homogeneous combustion reactions. The inhibition occurred most probably via free radical recombination at the solids surface. Vlachos et al. (1994) reported that the operating conditions and spatial heterogeneity (presence of solid surface) could largely affect product distributions. For example, removal of H atoms and CH3 radicals by adsorption can increase the ignition temperature. The inhibition effect from sand particles was observed in many studies [10, 35].
(3) T > 1150 K: Above this high temperature, methane conversion increased rapidly with increasing temperature. Moreover, the rate of increase of methane conversion was similar to the rate observed in the homogeneous reactor. At high temperatures, the free radical generation dominated the free radical quenching process. Hesketh et al. [10] made a similar observation in an incipient fluidized bed of sand.
The temperature required to achieve complete methane conversion depends on the average gas residence time inside the fixed bed reactor. Figure 2.1 shows that for an average gas residence time (τ) of 3 s, a reactor temperature of 1170 K was necessary. Figure 2.2 reports the methane conversion as a function of reactor temperature for τ = 6, 2 and 1.2 s. The required temperature to obtain complete conversion increased from 1130 K to 1190 K as the gas residence time dropped from 6 s to 1.2 s, respectively.
Fuel lean methane combustion in a fixed bed of sand particles can be modeled as a first-order single-step reaction. Sotudeh-Gharebagh et al. [31] showed that the methane conversion is
198
independent of the initial methane mole fraction in the range of 2-4 mol%. Therefore, the overall methane conversion is first order with respect to methane concentration.
2.2.1.1 Effect of particle size distribution and alumina particles
The effects of particle size distribution have been investigated by Sotudeh-Gharebagh et al. [31] in a fixed bed of sand particles of three mean sizes (dp = 230, 330 and 523 µm). The contact time index (h) was kept constant for all cases and particle size had no effect on conversion.
Sotudeh-Gharebagh et al. [31] also studied methane combustion in a fixed bed of alumina particles (dp = 363 µm) and they observed that alumina particles had the same effect as sand particles.
2.2.1.2 CO production
Methane combustion in a fixed bed of sand particles can lower CO compared to homogeneous combustion. Figure 2.3 shows the CO mole fraction in the off-gas of a fixed bed as a function of reactor temperature: a maximum in CO was observed near T = 1150 K. For hydrocarbon combustion, the formation rate of CO is greater than its consumption rate at low temperatures. As the temperature is increased, the CO consumption rate dominates and this explains the presence of a maximum in CO emissions.
In Figure 2.3, the CO measurements are compared to the predictions of the full and the reduced GRI mechanisms (Table 2.1) for homogeneous combustion. The measured CO mole fractions were low compared to the homogeneous model predictions for T < 1150 K. Sotudeh-Gharebagh et al. [31] attributed this effect to the low concentration of OH radicals, which were substantially reduced by the recombination reactions at the particle surface.
2.2.2 Kinetics for methane combustion in inert particles
Figures 2.1 to 2.3 compare the methane conversion in a fixed bed of sand particles measured by Sotudeh-Gharebagh et al. [31] with the predictions of the full and reduced GRI mechanisms (Table 2.1) for homogeneous combustion. The homogeneous models failed to predict the methane conversion and CO outlet mole fraction as a function of reactor temperature for an empty reactor (Figure 2.1) and for a fixed bed of sand (Figures 2.1 to 2.3) in the second temperature interval (1020 K < T < 1150 K). The discrepancy between the results in the empty reactor (homogeneous system) and the model may be due to the alumina wall, which is known to have an inhibitive effect comparable to sand, as previously discussed. To better predict combustion in the presence of solid particles, homogeneous models need to be modified in order to account for the heterogeneous steps. In the third temperature interval (T > 1150 K), where complete conversion was achieved, good agreement was observed between fixed bed experiments and model predictions. At these high temperatures, radical generation dominated the inhibition mechanism.
199
As previously discussed, methane conversion during fuel-lean catalytic combustion can be modeled using a single-step first-order model. However, a more complex model is required to characterize the production of intermediate species (CO, NOX, etc). Sotudeh-Gharebagh et al. [31] chose a certain number of reactions in their reduced mechanism (Table 2.1) and modified their reaction rates to represent the heterogeneous effects. The following assumptions were made:
• Free radical quenching occurs at the wall and particle surface [32, 36] • The recombination of OH, O, H and CH3 radicals retards homogeneous ignition [32] • In the presence of inert particles, the most important radicals are OH and H radicals.
Simulations with the reduced GRI mechanism (Table 2.1) showed that OH radicals play a major role in the inhibition process for the experimental conditions of Sotudeh-Gharebagh et al. [31].
Hence, the reaction rates (frequency factor and activation energy) of elementary steps #1, 2, 6, 12, 14, 16, 26, 27, 29 and 34 were modified in the reduced mechanism. They were changed based on a global reaction factor (f), which was defined as the ratio of the heterogeneous reaction rate to the homogeneous reaction rate:
−==
TR
∆EA
r
rf ahet exp
hom
(f < 1) (2.6)
The optimal values for the constant A and the activation energy (Ea) were found to be A = 4.3×109 and Ea = 217.2 kJ/mol, respectively. At temperatures above 1170 K, the model returns to the homogeneous case (f = 1). The activation energy is of the order of magnitude found in the combustion literature [37]. The activation energies for the modified mechanism are presented in Table 2.1.
In Figures 2.3 and 2.4, the calculated methane conversion and CO mole fractions as a function of reactor temperature are compared to the experimental measurements of Sotudeh-Gharebagh et al. [31] obtained in a fixed bed of sand. The modified kinetics fits the experimental data well in all temperature intervals contrary to the reduced homogeneous GRI model.
2.2.3 Combustion catalysts
Combustion catalysts are generally in the form of solid porous micro-spheres that promote heterogeneous reactions and reduce the lower flammability limit below 5 mol% (in air). Low temperatures (in the range of 773 K) and excess oxygen can result in zero emissions of NOX and CO. However, catalysts increase the operating costs for power generation. The kinetics of fuel-lean methane catalytic combustion has been reported as a first-order single-step mechanism [17,
200
38], which agrees with the observations of Sotudeh-Gharebagh et al. [31] with inert particles. More complex models are unnecessary since breakthrough of CO and NOX are zero.
2.2.4 Oxygen carriers
In chemical-looping combustion, a solid oxygen carrier is used to bring the oxygen to the fuel for reaction. Oxygen carriers are composed of an active metal oxide (ZnO, CuO and Fe2O3, for example) and an inert support/binder. To fully describe chemical-looping combustion, both reduction and oxidation steps are required. Reactions schemes have been reviewed by Hossain et al. [2] – the solids act as a catalyst in some reactions and as a reactant in others.
3 MIXING IN GAS-SOLID FLUIDIZED BEDS
The complex gas/solid hydrodynamics in fluidized beds may have a dominant influence on the prevailing mechanisms and rate determining steps during combustion since the reaction rates are fast compared to the mass transfer rates. Dispersion and contacting of the gas and solids depend on the particle properties, operating conditions, reactor scale and geometry.
3.1 Superficial gas velocity and bubble size
In the bubbling fluidization regime, the bubbles produce gas by-passing that limits the gas/solid mixing. The bubble-emulsion inter-phase mass transfer decreases with increasing superficial velocity and increasing bubble size [39]. In the turbulent fluidization regime, at high superficial gas velocity, gas/gas and gas/solid contacting is significantly improved [40].
3.2 Particle size
The addition of catalyst fines and the use of a wider particle size distribution are critical in maintaining a good quality of fluidization and are often sited as a prime factor to optimize reactor performance [41]. Furthermore, many experimental investigations show that the addition of catalyst fines in the bed leads to significantly higher global conversion [42, 43]. Fines affect the reactor by reducing the mass transfer limitations, since the intrinsic properties of the catalyst remain unchanged [44].
A reported effect of the addition of fines on fluid bed hydrodynamics is an increase in the bed voidage. This was shown by Rowe et al. [45], who used an X-ray absorption technique to measure the change in voidage of the emulsion phase following the addition of fines. This was also observed by Pell et al. [44]. In essence, this results in a higher level of permeability of the emulsion phase and a more homogeneous fluidization. Hence, more gas flows through the solid-rich emulsion phase to the detriment of the bubble phase, resulting in smaller bubbles and a higher inter-phase mass transfer. This was concluded by several models [43, 45]. Smaller fractions of gas into bubbles should result in a lower degree of by-passing and gas phase hydrodynamics closer to plug flow. Some experimental investigations have also reported that fines can stay suspended in bubbles [46, 47].
201
The mean particle size also affects gas/solid mixing in fluidized beds. In the turbulent regime, Sotudeh-Gharebaagh et al. [40] observed better gas mixing (Ug = 0.8 m/s) with FCC (dp = 70 µm) compared to sand (dp = 543 µm). On the other hand, Wu et al. [14] reported a higher bubble-emulsion mass transfer in a bubbling fluidized bed close to incipient fluidization conditions (Ug ~ Umf). Larger particles increase the minimum bubbling velocity (Umb) and the transition to turbulent fluidization (Uc), which may partially explain these results.
3.3 Effect of baffles
Another means of reducing the divergence from plug flow of a bubbling fluidized bed may reside in the utilization of reactor internals. Reactor internals are often necessary for heat transfer purposes and several studies have suggested that they could be an effective means for improving gas-solid contacting; the bubble size distribution can be homogenized and reduced, such that the the solid and gas axial circulation rate can be lowered [4, 48, 49]. The goal is to promote bubble splitting and gas redistribution into the emulsion phase [50]. However, there are limitations as Volk et al. [51] have shown that the axial solid mixing can be reduced to the level at which segregation is promoted and the fluidization of each individual compartment becomes difficult. Furthermore, the effectiveness of internals greatly depends upon its design [52].
3.4 Sparger flow pattern for non-premixed operation
The injection gas velocity at the sparger nozzle may lead to a number of patterns ranging from the periodic formation of bubbles to the formation of a permanent jet. At relatively low gas injection velocities, gas bubbles form, grow, detach and travel through the bed. As the injection velocity increases, bubbles grow bigger and the distance between successive bubbles decreases. At higher gas injection velocities, bubbles coalesce to form a plume or a jet at the sparger tip. The flow pattern and jet penetration length, Lj, defined as the distance between the plate and centre of the bubble at the instant it detaches from the jet, are important for process design since:
• The gas injected from spargers entrains solids in the vicinity of the nozzle, which causes
erosion on impinged surfaces and attrition • Knowledge of the jet flow could help maximize gas-solid contact and minimize attrition • For combustion processes, the reaction products largely depend on how the feed enters
the bed
A vast body of experimental data is available in the literature for vertically upward gas injection into fluidized bed reactors. Correlations for jet length are available for various fluidized bed operating conditions [47, 53, 54, 55, 56, 57]. Significantly less information is available for downward facing nozzles [40, 58]. Criteria to determine the flow pattern (bubbling or jetting) at the tip of an upward facing sparger have been proposed by Grace et al. [59] and Roach [46] as shown in Table 3.1.
202
The transition velocity (Umj) between the bubbling and jetting patterns can be measured experimentally by measuring the mean pressure fluctuations close to the sparger tip. Sotudeh-Gharebaagh et al. [40] measured a value of Umj = 115 m/s with a downward facing sparger (dor = 2 mm) inside a high temperature (Tb = 713 K) bubbling fluidized bed of sand particles and FCC catalyst. The criteria of Table 3.1 failed to correctly predict the flow pattern at the sparger tip most likely since they fall short of considering all reactor operating conditions (temperature, jet angle, injection velocity) and characteristics (reactor diameter and geometry). The formation of a jet is more likely with [59, 60]:
• Coarser and lighter particles • Increasing fluid density • Increasing orifice velocity • Low superficial velocity • Decreasing temperature • Jets away from the vertical (attrition by downward jets is significantly higher than that of
the upward jet [61]
In commercial fluidized bed reactors, where gas can be injected through a sparger with thousands of nozzles, the sparger is designed to operate under bubbling conditions (Uj = 30 m/s) to keep attrition low [6]. However, a jetting flow pattern at the sparger nozzle produces enhanced mixing compared to the bubbling pattern. Figure 3.1 shows the radial profiles of a non-reacting propane tracer at various axial positions above a downward facing sparger located at the centre of a fluidized bed [11]. At low gas injection velocity (Uj = 14.8 m/s) the tracer radial dispersion was poor - the tracer mole fractions reached a maximum at the bed centreline and decreased closer to the wall as shown in Figure 3.1(a). On the other hand, at a high gas injection velocity (Uj = 92.6 m/s), gas dispersion was enhanced as the tracer mole fraction was more uniformly distributed along the bed radius (Figure 3.1(b)). At low injection velocities, the tracer is mostly injected in bubbles that result in low gas-gas and gas-solid contacting.
The presence of the sparger may also limit mixing in its surrounding [40]. Figure 3.2 shows the methane concentration standard deviation as a function of axial position. Methane was injected from a downward facing sparger in a high temperature turbulent fluidized bed (Tb = 673 K and Ug = 0.8 m/s) and the standard deviation was calculated from the radial concentration profile. The sparger tip was located 170 mm above the distributor, which corresponds to Z = 0 mm on Figure 3.2. The data from two spargers are presented: a non-baffled sparger and a baffled sparger. The high standard deviation close to the spargers can be attributed to the fact that bubbles formed at the sparger nozzles and then rose close to the sparger wall. The wall acted as a barrier to mixing as it reduced contact between the sparger bubbles and the gas and solids inside the fluidized bed. As soon as the bubbles went past the sparger level, the tracer gas was well dispersed to the emulsion phase, due to intense gas-solid and gas-gas mixing in the turbulent fluidization regime. The standard deviation dropped significantly within less than 100 mm above the sparger. Figure 3.2 also shows the existence of three distinct mixing regions: null, partial and complete mixing. In the null mixing zone located below the sparger (Z < 0 mm), tracer gas is absent. In the acceleration zone (0 ≤ Z ≤ 25 mm), the axial and radial gas mixing is limited by the
203
sparger-nozzle walls. In the fully developed mixing zone (Z > 25 mm), above the tracer injection level, a fairly uniform mixing pattern is developed.
Figure 3.2 also shows that baffled spargers can enhance mixing compared to non-baffled spargers. At 2 mm below the sparger tip, for baffled spargers, the gas diffusion was deeper as compared to the non-baffled sparger and the equilibrium methane concentration was reached faster with the baffled sparger (Z = 200 mm) compared to the non-baffled sparger (Z = 300 mm).
4. METHANE FLUIDIZED BED COMBUSTION
Fluidized bed combustion is characterized by both kinetics and gas/solid hydrodynamics. Combustion can occur in different regions inside the fluidized bed (at the distributor, at the sparger tip, in the bubble/emulsion phases and in the freeboard region), which has a significant effect on fuel conversion and pollutant emissions. Combustion of gaseous mixtures, mostly methane-air, has been successfully implemented in fluidized bed reactors.
4.1 Inert Particles
Most studies on fluidized bed combustion with inert particles have been performed in premixed mode [7, 8, 9, 10, 15]. Few were achieved with fuel injection through a separate sparger [11, 12, 13].
4.1.1 Fluidized bed temperature
The fluidized bed temperature is a key parameter to ensure complete combustion. With increasing temperature, the combustion is initiated deeper inside the fluidized bed. In premixed combustion, increasing temperature moves the combustion front towards the distributor - the combustion front moves first from the freeboard to the bubbles bursting at the upper surface of the sand, then to bubbles igniting in the bed and finally to small bubbles where ignition occurs practically at the level of the distributor and the process appears flameless [9, 35]. Two critical transition temperatures have been defined for fluidized bed combustion. Below the lower critical temperature (T1), combustion only occurs above the bed surface. Between T1 and the upper critical temperature (T2), combustion begins to move inside the bubbles within the bed. Finally, above T2, combustion takes place entirely within the bed and close to the distributor.
Table 4.1 summarizes methane fluidized bed combustion studies that measured T1 and/or T2. The values of T1 and T2 reported vary by 100 - 200 K. This difference may be due to the difference in operating conditions and the method by which the critical temperatures were determined – temperature or species measurements. Van der Vaart [62] reported that T1 increased with decreasing bed height, increasing excess fluidizing gas velocity and increasing particle size.
Combustion will occur inside the emulsion (particulate) phase if the fluidized bed temperature is sufficiently high above T1. Hesketh et al. [10] observed in-bed combustion inside an incipient
204
fluidized bed for Tb ≥ 1188 K. Sotudeh-Gharebagh et al. [31] observed complete methane conversion in a fixed bed of silica sand at Tb ≥ 1150 K. At sufficiently high temperature, free radicals production outweighs the radical recombination at the surface of particles and combustion is possible in the emulsion phase.
In non-premixed combustion, increasing the temperature will bring the combustion front closer to the sparger. The sparger operating conditions and design should be chosen to maximize gas/gas and gas/solid contacting.
4.1.2 Bubble size and air-fuel ratio
Friedman et al. [8] reported that increasing the air-fuel ratio promotes the consumption of methane in the bed at lower bed temperatures (below 1173 K). Wu et al. [14] injected bubbles of pure methane (non-premixed combustion) inside a bubbling fluidized bed of alumina and sand particles. It was found that small bubbles enhanced bubble-emulsion mass transfer, which promoted oxygen diffusion inside the bubbles and bubble-phase combustion. Wu et al. [14] also reported that increasing the oxygen concentration in the emulsion promoted bubble-phase combustion.
Pilawska et al. [64] observed that when Tb < T1 (1123 K in their case) only bubbles above a certain critical size ignite on leaving the fluidized bed. Complete methane combustion was only observed when the superficial velocity was sufficiently higher (Ug/Umf > 4 for a bed height of 20 mm).
4.1.3 Fluidization regime
The turbulent fluidized bed regime shows better gas-gas and gas-solid contacting compared to the bubbling regime. Also, due to its higher throughput, the turbulent regime offers a higher power generation for the same operating temperature. The power generated in the combustion process is calculated using the following expression:
HXFP CH ∆= 4 (4.1)
4.1.4 CO formation
The combustion of hydrocarbons can be viewed as a two-step global process: combustion to CO and its conversion to CO2. At lower temperatures and also in the presence of inert particles, consumption of CO is very small compared to its formation and conversely at higher temperatures. The bed temperature must be sufficiently high to completely oxidize CO to CO2 within the bed.
Sotudeh-Gharebaagh et al. [65] showed that the conversion of CO to CO2 occurs faster in a turbulent fluidized bed. Pure methane was injected through a downward facing sparger at two
205
injection gas velocities (Uj = 30, 120 m/s) at a bed temperature of 1173 K. These sparger velocities represented two different sparger hydrodynamics: bubbling and jetting. Furthermore, two gas superficial velocities (Ug = 0.7, 1.5 m/s) were tested, which corresponded to the bubbling and turbulent regimes, respectively.
In Figure 4.1, CO measurements along the reactor height at the four conditions are reported (the sparger tip was located at Z = 170 mm). For all cases, the CO volume fraction increased close to the sparger tip where combustion occurred under fuel-rich conditions due to incomplete gas mixing in that region. The CO volume fraction then reached a maximum value downstream and then declined as the gases mixed and reacted further with O2 in the fluidized bed. All CO was converted to CO2 since the bed temperature was sufficiently high.
As shown in Figure 4.1, increasing gas injection velocities resulted in greater CO peaks due to enhanced mixing at the sparger tip, which led to faster combustion and increased dispersion of gas in the emulsion phase. The complete reaction length for CO oxidation in the turbulent fluidization was also small compared to bubbling conditions due to improved gas mixing. The overall trend reported in Figure 4.1 is in close agreement with the predictions as well as with the experimental and theoretical findings of turbulent non-premixed combustion of methane in air [66] and of coal combustion [67].
Figure 4.2(a) shows the normalized concentration of CO (ppm) measured at the reactor outlet for non-premixed methane combustion in a bubbling and turbulent fluidized bed. For bed temperatures less than about 1123 K, high concentrations of CO were generated since the temperature was insufficient to completely convert the CO within the bed. In the temperature range of 1123-1273 K, both systems performed quite well. Turbulent fluidized beds generated less CO most probably due to higher contacting efficiency, while bubbling fluidized beds need to operate above 1170 K for low CO emission levels. This agrees with the observations of Friedman et al. [8]. Furthermore, Friedman et al. [8] reported that increasing gas superficial velocity in the bubbling fluidization regime reduced in-bed CO conversion because of increasing bubble size, which resulted in higher bubble rise velocities and gas by-passing.
4.1.5 NOX formation
A major environmental concern for fluidized bed systems is to control NOX emissions. NOX formation and destruction in combustion processes result from a combination of fuel nitrogen oxidation, thermal processes, and reactor hydrodynamics. The success of natural gas combustors depends strongly on their operating temperature, which is critical for thermal NOX emissions. Thermal NOX are generated in conventional devices where the flame temperature is usually above 1473 K. Below 1073 K, no thermal NOX is produced.
Fluidized beds are characterized by solids mixing that prevent the formation of hot spots, which reduces the NOX compared to homogeneous gas-phase combustion. Friedman et al. [8] reported that NOX emissions were considerably reduced when the combustion front moved from the
206
freeboard to the fluidized bed. NOX emissions have been observed to increase with bed temperature until a critical temperature where the trend reverses. This critical temperature was measured as ~1100 K and 1200 K by Friedman et al. [8] and Pilawska et al. [64], respectively.
Friedman et al. [8] reported that increasing the gas superficial velocity (constant air-fuel ratio) increased NOX emissions. It was argued that a denser bed could more easily quench hot spots. Pilawska et al. [64] observed an increase in NOX with increasing Ug/Umf. In this case, it was hypothesized that bigger bubbles resulted in enhanced gas by-passing and less in-bed combustion.
Sotudeh-Gharebaagh et al. [65] compared the NOX emissions from a bubbling and a turbulent fluidized bed of sand particles. In Figure 4.2(b), NOX fractions are reported for turbulent and bubbling conditions. The NOX emission levels for turbulent fluidized-bed operation were higher compared to the bubbling bed. This was attributed to the higher availability of oxygen in the turbulent fluidized bed, which was almost 8% above stoichiometry. In comparison, the bubbling bed was operated with excess oxygen of less than 1%, which consequently led to lower NOX formation. The experimental NOX measurements in Figure 4.2(b) were significantly lower than the predictions of the GRI model for homogeneous combustion. For industrial reactors operating in the turbulent fluidization regime with an integrated heat exchanger inside of the reactor and less oxygen, NOX emissions can be significantly lowered. NOX emissions as low as 5 ppm were measured by Friedman et al. [8] and Pilawska et al. [64] in the bubbling regime with inert particles.
4.1.6 Particle type and size
Sotudeh et al. [40] reported that the CO generated in a turbulent fluidized bed of sand particles (~ 10 mol%) was greater compared to FCC particles (~ 2 mol%). Gas/gas and gas/solid mixing is significantly improved with small FCC particles (dp = 70 µm) compared to large sand particles (dp = 543 µm). With sand particles, limited mixing resulted in fuel-rich conditions and reactions that were mostly homogeneous.
4.2 Combustion catalysts
Studies have been performed with various combustion catalysts, which have shown that low temperature combustion (Tb ~ 773 K) of a fuel lean methane-air mixture can be achieved with zero CO and NOX emissions in a turbulent fluidized bed [17, 38].
4.3 Oxygen carriers / chemical-looping combustion
Chemical-looping combustion is typically done in a circulating fluidized bed. The concept is based on shuttling catalyst between two zones: the oxygen carrier is reduced in a transport bed/riser and then re-oxidized in a fluidized bed regenerator [2].
207
5 FLUIDIZED BED COMBUSTION MODELLING
Methane conversion in gas-solid fluidized bed has been observed to vary from plug flow to backmixed flow, depending on the reaction and fluidization conditions. Various models have been proposed to describe the performance of fluidized beds and these models differ greatly in complexity.
Empirical models are seldom of much value due to the hydrodynamic complexity of fluidized beds and the large number of variables involved. Historically, mechanistic models have been based on a pseudo-homogeneous or a two-phase approach. The pseudo-homogeneous approach describes the fluidized bed hydrodynamics with single-phase models (well-mixed, plug flow and dispersion models), which constitutes a poor characterization.
In the two-phase approach, the fluidized bed is separated into two one-dimensional paths: one of high voidage that represents the bubble phase and one of high solids fraction that represents the dense (emulsion or particulate) phase. The mole balance is written for both phases and includes terms for the chemical reactions and the inter-phase mass transfer. The performance of two-phase models is dependent upon the inter-phase mass transfer, the volume of each phase, the solids fraction the chemical kinetics and the flow rate attributed to each phase. Two-phase models, reviewed in several studies [6, 68, 69, 70], are widely used and have been quite successful for steady-state fluidized bed gas-phase solid-catalyzed reactions.
5.1 Single-phase models
Single-phase models are reviewed in several studies [6, 69]. Lorences et al. [4] showed that for Ug ≤ 6 Umf, the gas hydrodynamics is very close to plug flow such that the conversion for a first order reaction is within 0.8% of the value for plug flow.
5.2 Two-phase models
Two-phase models describe the fluidized bed with a molar conservation equation for the bubble and emulsion phases. The conservation equations contain an inter-phase transfer term and the two phases are generally modeled as ideal reactors (plug flow or CSTR). Several fluidized bed models have been published in the literature and include the model of Davidson and Harrison [71], bubbling bed models [6] and bubble assemblage models [72].
5.2.1 The model of Davidson and Harrison
General two-phase models have the following assumption [71]:
• All gas flow in excess of Umf passes through the bed as bubbles • Bubbles are of uniform size throughout the bed
208
• Interphase mass transfer occurs through a combined effect of molecular diffusion and gas throughflow
• The bubble and emulsion phases are either perfectly mixed or in plug flow
5.2.2 Bubbling bed models
The bubbling bed model of Kunii and Levenspiel [6] is a modified version of the two-phase model where a cloud-wake phase is also considered. Hence, the interphase mass transfer occurs between the bubble and the cloud-wake phase and between the later and the emulsion phases.
5.2.3 Bubble assemblage models
Other hydrodynamic models have used various combinations of assumptions. Among them, the Kato and Wen [72] model considers the change of bubble size with height inside the bed. Modified versions of this model have also been proposed by Dounit et al. [7].
5.2.2 Multiple region models
The grid region near the distributor, the region surrounding the tip of spargers and the freeboard zone above the fluidized bed are characterized by hydrodynamics significantly different from the main body of the fluidized bed. These regions are characterized by relatively high voidage due to the injection of gas (grid and sparger tip) and the disengagement of solids (freeboard).
5.2.2.1 Models for the grid region
A significant amount of reaction may occur in the grid region, especially for fast reactions such as combustion [73]. The jet of gas originating from the distributor results in good contacting. Behie et al. [74] calculated an inter-phase mass transfer coefficient 40 to 60 times higher in the grid region compared to the main body of the fluidized bed. Single-phase models have been proposed to describe the grid region [73, 75, 76]. Behie et al. [74], Grace et al. [77] and Sit et al. [78] proposed two-phase models with enhanced inter-phase mass transfer terms.
5.2.2.2 Models for the freeboard region
In the freeboard region, solid particles are projected due to the bubbles bursting at the bed surface. Reaction can occur in the freeboard region due to the high voidage. Studies have shown that the solids fraction decreases exponentially with height from the bed surface and reaches a constant value above the transport disengagement height (TDH) [79, 80]. Yates et al. [81] developed a single-phase model based on the assumption that the freeboard contained perfectly mixed particles derived from bubble wakes. Chen et al. [82] modeled the freeboard with an axial dispersion model and an exponential decrease in solids hold-up.
5.3 Kinetic models
209
Fluidized beds are characterized by complex gas/solid hydrodynamics, such that simple kinetic models are convenient. Methane conversion during fluidized bed combustion with inert particles has been modeled with global homogeneous models. Pre et al. [83] modelled high temperature (Tb ≥ 973 K) methane combustion in a fluidized bed of silica sand particles with the two-step model of Dryer et al. [19], Bradley et al. [20] and Westbrook et al. [18]. The kinetics of Dryer et al. [19] fitted more closely the experimental data. Dounit et al. [7, 84] also modeled methane combustion in a fluidized bed of silica sand with reasonable agreement with the experimental data. Wu et al. [14] compared the three models and found that the model of Bradley et al. [20] reproduced their experimental fluidized bed combustion data more closely. This emphasizes the fact that global models are more accurate for the conditions for which they were developed. Furthermore, more detailed models are required to describe the formation of intermediates (CO, NOX), such as the corrected model in Table 2.1 [85].
Notation
A Constant (Eq. 2.6)
C Concentration (kmol/m3)
dj Nozzle diameter (m)
dp Mean particle diameter (µm)
dor Orifice diameter (m)
Ea Activation energy (kJ/mol)
f Ratio of heterogeneous to homogeneous reaction rates
F Molar flow rate (mol/s)
g Gravitational acceleration (9.81 m2/s)
h Contact time index (g s/mol)
∆H Heat of reaction (kJ/mol)
k Reaction rate constant
Lj Characteristic jet length (mm)
n Temperature exponent
P Power (kW)
R Universal gas constant (8.314 J/mol K)
r Reaction rate (kmol/m3 s)
Ug Superficial gas velocity (m/s)
Uc Gas velocity at the onset of turbulent fluidization (m/s)
Umb Minimum bubbling velocity (m/s)
Umf Minimum fluidization velocity (m/s)
Umj Minimum jetting velocity (m/s)
210
Uj Gas discharge velocity at nozzle tip (m/s)
T Temperature (K)
Tb Bed temperature (K)
T1 Lower critical temperature (K)
T2 Upper critical temperature (K)
X Conversion (%)
Y Molar fraction (mol%)
Z Axial position (mm)
Greek symbols
δ Threshold value
ε Void fraction
φ Fuel-air ratio
ρ Density (kg/m3)
τ Average gas residence time (s)
Subscript
a Refers to alumina particles
corr Corrected
f Gas-phase
het Heterogeneous
hom Homogeneous
i ith reaction
max Maximum
s Solids
0 Conditions at the inlet
References
[1] R.E. Hayes, S.T. Kolaczkowski, Introduction to catalytic combustion, Gordon and Breach Science Publishers, The Netherlands, 1997.
[2] M.M. Hossain, H.I. de Lasa, Chem. Eng. Sci. 2008, 63, 4433-4451.
[3] J.R. Grace, Gas Fluidization Technology, Chapter 11, Wiley, Chichester, U.K., 1986.
[4] M.J. Lorences, J-P Laviolette, G.S. Patience, M. Alonso, F.V. Diez, Powder Technol. 2006, 168, 1-9.
[5] B. Du, L.-S. Fan, F. Wei, W. Warsito, AIChE J. 2002, 48, 1896-1909.
211
[6] D. Kunii, O. Lenvenspiel, Fluidization Engineering, 2nd ed., Butterworth-Heinemann, Stoneham, MA, USA, 1991.
[7] S. Dounit, M. Hemati, R. Andreux, Chem. Eng. J. 2008, 140, 457.
[8] J. Friedman, H. Li, Combust. Sci. Technol. 2005, 177, 2211-2241.
[9] W. Żukowski, Combust. Flame 2003, 134, 399-409.
[10] R.P. Hesketh, J.F. Davidson, Combust. Flame 1991, 85, 449-467.
[11] R. Mabrouk, J. Chaouki, J-P Laviolette, G.S. Patience, Proc. 9th Conference on Circulating Fluidized Beds, Hamburg, Germany, 2008.
[12] R. Sotudeh-Gharebaagh, PhD Thesis, Ecole Polytechnique, Montreal, Canada, 1998.
[13] B. Singh, G.R. Rigby, T.G. Callcott, Inst. Fuel Symp. Series 1: Fluidised Combustion, C5-1, 1975.
[14] W. Wu, P.A. Dellenback, P.K. Agarwal, H.W. Haynes Jr., Combust. Flame 2005, 140, 204-221.
[15] R.A. Srinivasan, S. Sriramulu, S. Kulasekaran, P.K. Agarwal, Fuel 1998, 77, 1033-1049.
[16] J.R. Grace, Chem. Eng. Sci. 1990, 45, 1953.
[17] M. Foka, J. Chaouki, C. Guy, D. Klvana, Chem. Eng. Sci. 1994, 49 (24A), 4269-4276.
[18] C.K. Westbrook, F.L. Dryer, Combust. Sci. Tech. 1981, 27, 31.
[19] F.L. Dryer, I. Glassman, Proc. 14th Symp. (Intl.) on Combust. 1973, 987.
[20] D. Bradley, S.B. Chin, M.S. Draper, G. Hankinson, Proc. 16th Symp. (Intl.) on Combust. 1977, 1571.
[21] W.C. Gardiner, Combustion Chemistry, Springer-Verlag, New York, 1984.
[22] D.R. Van der Vaart, Ind. Eng. Chem. Res. 1992, 31, 999-1007.
[23] URL: http://www.chem.leeds.ac.uk/Combustion/Combustion.html
[24] URL:http://ame-www.usc.edu/research/combustion/combustionkinetics/research.html
[25] URL: http://www.me.berkeley.edu/gri-mech/
[26] A.S. Tomlin, T. Turányi, M.J. Pilling, Low Temperature Combustion and Autoignition, Elsevier, Amsterdam, 1997.
[27] J. Warnatz, U. Mass, R.W. Dibble, Combustion, Springler-Verlag, New York, 1999.
[28] M. Frenklach, Numerical Approaches to Combustion Modeling, Progress in Astronautics and Aeronautics, AIAA, Washington, 135, 1991, 129-154.
[29] J. Warnatz, Proc. 24th Symp. (Intl.) on Combust. 1992, 553.
[30] W.C. Chang, J.Y. Cheng, Reduced Mechanisms Based on GRI-Mech 1.2, http://www.me.berkeley.edu/drm/.
[31] R. Sotudeh-Gharebagh, J. Chaouki, Can. J. Chem. Eng. 2003, 81, 1182-1191.
[32] D.G. Vlachos, L.D. Schmidt, R. Aris, AIChE J. 1994, 40, 1018-1025.
212
[33] R.J. Kee, F.M. Rupley, J.A. Miller, SAND89-8009 Sandia National Laboratories Report, CA, USA, 1989.
[34] D.J. Tringham, PhD Thesis, Leeds, U.K. 1982.
[35] J.S. Dennis J.S., A.N. Hayhurst, I.G. Mackley, 19th Symp. (Intl.) on Combust. 1982, p. 1205.
[36] B. Lewis, G. Von Elbe, Combustion, Flames and Explosions of Gases, 3rd ed., Academic Press, London, 1987.
[37] P. Markatou, L.D. Pfefferle, M.D. Smooke, Combust. Flame 1993, 93, 185-201.
[38] J. Chaouki, D. Klvana, C. Guy, Korean J. Chem. Eng. 1999, 16 (4), 494-500.
[39] W. Wu, P.K. Agarwal, Can. J. Chem. Eng. 2003, 81, 940-948.
[40] R. Sotudeh-Gharebaagh, J. Chaouki, Can. J. Chem. Eng. 2000, 78, 65-74.
[41] J.R. Grace, G. Sun, Can. J. Chem. Eng. 1991, 69, 1126-1134.
[42] S.C. Arnold, M. Cecchini, G. Fenati, Fluidization X, Engineering Foundation, New York, 181-188, 2001.
[43] J.G. Yates, D. Newton, Chem. Eng. Sci. 1986, 41, 801-806.
[44] M. Pell, S.P. Jordan, Fluidization Engineering: Fundamentals and Applications 1988, 84, 68-73.
[45] P.N, Rowe, L. Santoro, J.G. Yates, Chem. Eng. Sci. 1978, 33, 133-140.
[46] P.E. Roach, Int. J. Multiphase Flow 1993, 19, 1159-1165.
[47] W.C. Yang, B. Ettehadieh, T.C. Anestis, R.E. Gizzie, G.B. Haldipur, Fluidization V, Engineering Foundation, New York, 95-102, 1986.
[48] H.V. Nguyen, O.E. Potter, A.B. Whitehead, Chem. Eng. Sci. 1979, 34, 1163-1164.
[49] S. Ramamoorthy, N. Subramanian, Chem. Eng. J. 1981, 22, 237-42.
[50] N. Dropka, K. Jaehnisch, E.V. Kondratenko, V. Kondratenko, R. Kraehnert, N. Steinfeldt, Int. J. Chem. React. Eng. 2005, 3, 1-16.
[51] W. Volk, C.A. Johnson, H.H. Stotler, Chem. Eng. Prog. 1962, 58, 47-44.
[52] S. Dutta, G.D. Suciu, J. Chem. Eng. Jpn. 1992, 25, 345-348.
[53] J.G. Yates, V. Bejcek, D.J. Cheesman, Fluidization V, Engineering Foundation, New York, 79-86, 1986.
[54] J.M.D. Merry, AIChE J. 1975, 21, 507-510.
[55] C.Y. Wen, M. Horio, R. Kirshnan, R. Khosravi, P. Rengarajan, Proc. 2nd Pacific Chem. Eng. Cong., Nagoya, Japan, 1182-1189, 1977.
[56] T.R. Blake, H. Wevv, P.B. Sunderland, Chem. Eng. Sci. 1990, 45, 365-371.
[57] W.C. Yang, D.L. Keairns, Ind. Eng. Chem. Fund. 1979, 18, 317-320.
[58] Z. Shen, C.L. Briens, M. Kwauk, M.A. Bergougnou, Can. J. Chem. Eng. 1990, 68, 534-540.
[59] J.R. Grace, C.J. Lim, Can. J. Chem. Eng. 1987 65, 160.
213
[60] L. Massimilla, Fluidization, 2nd ed., Academic Press, New York, 1985.
[61] J. Werther, W. Xi, Powder Tech. 1993, 76, 39-46.
[62] D.R. Van der Vaart, Fuel 1988, 67, 1003-1007.
[63] R. Sotudeh-Gharebaagh, J. Chaouki, Energ. Fuel. 2007, 21, 2230-2237.
[64] M. Pilawska, C.J. Butler, A.N. Hayhurst, D.R. Chadeesingh, Combust. Flame 2001, 127, 2181-2193.
[65] R. Sotudeh-Gharebaagh, J. Chaouki, R. Legros, Chem. Eng. Sci. 1999, 54, 2029-2037.
[66] A.R. Masri, R.W. Bilger, R.W. Dibble, Combust. Flame 1988, 73, 261.
[67] J. Lufei, H.A. Becker, R.K. Code, Can. J. Chem. Eng. 1993, 71, 10.
[68] T.C. Ho, Handbook of Fluidization and Fluid-Particle Systems, Chapter 9, Taylor & Francis, USA, 2003.
[69] J.G. Yates, Fundamentals of Fluidized-Bed Chemical Processes, Butterworths, London, 1983.
[70] C.A. McKnight, L.P. Hackman, J.R. Grace, A. Macchi, D. Kiel, J. Tyler, Can. J. Chem. Eng. 2003, 81, 338–350.
[71] J.F. Davidson, D. Harrison, Fluidized Particles, Cambridge University Press, London, 1963.
[72] K. Kato, C.Y. Wen, Chem. Eng. Sci. 1969, 24, 1351-1369.
[73] J.G. Yates, P.N. Rowe, D.J. Cheesman, AIChE J. 1984, 30 (6), 890-894.
[74] L.A. Beheie, P. Kehoe, AIChE J. 1973, 19 (5), 1070-1072.
[75] P.N. Rowe, Chem. Eng. Sci. 1993, 48 (14), 2519-2524.
[76] Y. Xiao-tian, D.G. Horne, J.G. Yates, P.N. Rowe, AIChE Symposium Series 1984, 80 (241), 41-44.
[77] J.R. Grace, H.I. de Lasa, AIChE J. 1978, 24 (2), 364-365.
[78] S.P. Sit, J.R. Grace, Fluidization V, Engineering Foundation, USA, 39-46, 1986.
[79] C.Y. Wen, L.H. Chen, AIChE J. 1982, 28 (1), 117-128.
[80] D. Kunii, O. Levenspiel, Powder Technol. 1990, 61, 193-206.
[81] J.G. Yates, P.N. Rowe, Trans. Inst. Chem. Eng. 1977, 55, 137-142.
[82] L.H. Chen, C.Y. Wen, AIChE J. 1982, 28 (6), 1019-1027.
[83] P. Pré, M. Hemati, B. Marchand, Chem. Eng. Sci. 1998, 53 (16), 2871.
[84] S. Dounit, M. Hemati, D. Steinmetz, Powder Tech. 2001, 120, 49.
[85] R. Sotudeh-Gharebagh, J. Chaouki, P. Sauriol, Fuel Process. Technol. 2007, 88, 847-858.
214
ANNEXE 8 – ARTICLE PUBLIÉ DANS LE JOURNAL « APPLIED
CATALYSIS A: GENERAL » (2010)
La contribution à cet article se résume à l’écriture de la première version qui a été soumis aux co-
auteurs de DuPont pour révision. Ces derniers ont apportés plusieurs changements avant la
soumission au journal « Applied Catalysis A: General ». Il faut noter que les travaux
expérimentaux et l’analyse initiale des résultats ont été effectués chez DuPont. Cet article est
présenté dans cette thèse puisqu’il est pertinent et puisque l’auteur y a consacré plusieurs mois.
Parametric Study of n-Butane Oxidation in a Circulating Fluidized Bed Reactor
Manuscipt No. APCATA-D-09-01052
Keith W. Hutchensona*, Concetta La Marcab, Gregory S. Patiencec,
Jean-Philippe Laviolettec, Richard E. Bockratha
aDuPont Company, Experimental Station, Wilmington DE 19880, U.S.A. bDuPont Company, DuPont Engineering Research & Technology, DuPont Building, Wilmington
DE 19898, U.S.A. cDepartment of Chemical Engineering, École Polytechnique de Montréal, C.P. 6079, Succ.
“CV”, Montréal, QC, Canada, H3C 3A7
Submitted to Applied Catalysis A: General Special Issue on “International VPO Workshop”
Revision of
January 23, 2010
Keywords: Autoignition, n-Butane Oxidation, Circulating Fluidized Bed, Combustion, Induction Time, Maleic Anhydride, Oxidation, Vanadyl Pyrophosphate
∗To whom correspondence should be addressed. E-mail: [email protected]. Tel.: +1-302-695-1389, Fax: +1-302-695-2504
215
Abstract
DuPont’s n-butane oxidation process uses a proprietary attrition-resistant vanadyl pyrophosphate
(VPP) catalyst and circulating fluidized bed (CFB) technology for the selective oxidation of n-
butane to maleic anhydride. In the redox mode, n-butane feed plus recycle gas are fed to the riser
side of the CFB, and air is fed to the regenerator. To optimize both activity and selectivity,
oxygen may be fed to the riser side together with n-butane. These conditions fall within the n-
butane flammability envelope, and localized combustion reactions were experienced during
operation of a pilot plant utilizing this reactor technology. Complementary experimental and
theoretical investigations were conducted using a laboratory fluidized bed reactor and an
elementary step kinetic model to study the local reaction behavior near an oxygen sparger placed
in a fluidized bed of solid VPP catalyst and to evaluate the effect of operating conditions on the
onset of combustion in the homogeneous gas phase regions of the reactor. The laboratory
fluidized bed reactor was operated at temperatures of 320-390°C, pressures of 3-4 bar, freeboard
residence times of 4-8 s, and n-butane and oxygen compositions of 4-10% and 1-6%,
respectively. For the low solids regions of the reactor, both experimental and modeling results
demonstrated pronounced pressure and temperature effects on the onset of combustion in this
range of operating conditions and that decreasing the residence time delays the onset of complete
combustion to longer time or higher relative molecular oxygen compositions. Kinetic model
predictions also showed a significant influence of reactant composition and reactor pressure on
oxidation induction time.
INTRODUCTION
Many industrial catalytic processes involve simultaneous heterogeneous and homogeneous
reaction pathways. Homogeneous reactions generally result in yield loss of the desired product.
Moreover, they may also represent a significant safety hazard in the case of fast and highly
exothermic reactions such as those involving oxygen and hydrocarbons. Hydrocarbon selective
oxidation processes are widespread in the industry, and new processes are presently being
developed in the context of alkane functionalization. In fact, the direct oxidative conversion of
lower alkanes is undergoing intensive research due to their lower cost compared to olefins.
Fluidized bed reactors are currently being developed for a variety of these processes. To
216
maximize productivity, they may operate within the flammability limits while feeding the oxidant
and hydrocarbon separately into the bed—through spargers, for example. This is possible due to
the ability of the solids phase to suppress non-selective homogeneous reactions (combustion) and
at the same time to promote the selective heterogeneous reactions. Feed compositions to fixed
beds, on the other hand, are generally constrained by the flammability region. Mixing
hydrocarbons and oxygen upstream of the reactor is a major safety concern, and thus the optimal
feed composition may be dictated by flammability limits and not kinetics.
Oxidation processes may operate under partial reactant conversion because selectivity generally
declines at high conversion due to an increase in the rate of product decomposition [1-3]. The
combination of partial hydrocarbon conversion and high oxygen concentration results in
potentially flammable mixtures either within the catalyst bed or at the exit. In fluidized beds,
when the solids volume fraction is sufficiently high, free radicals may be quenched, thus reducing
the flammability risk. However, downstream of the catalyst bed in the freeboard, cyclones, and
associated piping, the risk of combustion is much greater since the effluent gas phase
composition is potentially flammable due to the high hydrocarbon and oxygen concentrations and
elevated temperatures, and the solids volume fraction may be insufficient to quench free radicals
and non-selective homogeneous reactions [4, 5]. Minimizing the risk of gas phase combustion
and deflagration in these regions remains an important design issue.
Maleic anhydride is currently produced in fixed and fluidized bed reactors via the selective
oxidation of n-butane with vanadyl pyrophosphate catalysts (VPP) [6]: feed compositions in the
fixed bed process are typically in the range of 2% n-butane in air, whereas the mixed cup
composition in the fluidized bed process exceeds 4% n-butane in air. Fixed beds are further
constrained to low hydrocarbon concentrations due to the poor heat transfer rates that may cause
hot spots, which may deactivate the catalyst [7].
Flammable mixtures may form either at the entrance of the catalyst bed (in the case of fixed
beds), in the bed of catalyst solids (in the bubble phase of fluidized beds, for example), or
downstream of the solids. The threshold conditions leading to autoignition (self-initiated
homogeneous combustion) must be known to minimize the safety hazards, not only during
standard operation, but also during start-up as well as for emergency shut-down conditions.
Several criteria to characterize a system’s sensitivity to autoignition include flammability limits,
217
minimum oxygen concentration (MOC), autoignition temperature (AIT), and induction time.
However, the induction time, defined as the minimum gas residence time that results in the onset
of homogeneous combustion, is the only criterion that yields the time-scale of the ignition
process. This information is key to maximizing the reactants’ concentration since it may allow
safe operation inside the flammability limits and above the AIT by ensuring that the induction
time is lower than the gas residence time.
Hydrocarbon/oxygen induction times are measured by rapidly heating and pressurizing a mixture
to a specific initial state, and then measuring the time it takes for the mixture to react (ignite).
Butane/oxygen induction times have been previously measured in several types of experimental
apparatus: static vessels, flow reactors [8], shock tubes [9], and high compression machines [10,
11]. A number of correlations have been developed relating the induction time to initial
operating variables including mixture composition, pressure, and temperature. However, n-
butane/oxygen induction times are unavailable in the temperature and pressure ranges of interest
for the n-butane selective oxidation process (T = 320–420°C and P = 2–4 bar). Low temperature
(≤ 450°C) gas-phase oxidation studies of n-butane have been performed at sub-atmospheric
pressure [12-14]. Furthermore, investigations at moderate pressure (1-6 bar) have only been
conducted at very high temperature (≥ 1000°C) [8, 15, 16]. Not only are the fluidized bed
operating conditions outside the range of these correlations, but the presence of catalyst solids
also affects the homogeneous reaction kinetics. Together with the quenching behavior due to the
catalyst volume, the nature of the catalyst surface may also affect the homogeneous kinetics in
different ways. Heterogeneous and homogeneous reactions are strongly coupled through heat
and mass transfer as well as through reactive intermediates that are common to both pathways.
Heterogeneous surface reactions can heat the catalyst boundary layer sufficiently to initiate
homogeneous reactions, can deplete the boundary layer of a limiting reactant, and can act as a
free radical source or sink. Previous studies have shown that flammability limits become
narrower on several noble metal catalysts, which is most likely a consequence of reactant
depletion in the catalyst boundary layer [17, 18]. Furthermore, Hesketh et al. [4] have shown that
inert sand particles strongly inhibit homogeneous reactions via radical recombination. The effects
of VPP catalyst on induction time have yet to be quantified.
This paper describes DuPont’s pilot and commercial plant experience regarding non-selective gas
phase oxidation reactions at gas spargers that were used to inject oxygen into a circulating
218
fluidized bed reactor (CFB) as well as the exit region of the CFB downstream of the cyclones.
This experience resulted from the process development and commercialization of a process to
manufacture tetrahydrofuran from n-butane. Scheme 1 shows the global reaction steps of the
overall process, including the selective oxidation of n-butane to maleic anhydride, which was the
reaction step conducted in this CFB reactor. This scheme also shows the alternative reaction
pathway of non-selective combustion of the n-butane. Together with detailed reaction
engineering modeling, an extensive parametric experimental program was developed to examine
the effects of temperature, pressure, gas phase composition, and catalyst concentration on the
phenomena. Some of these results are reported herein. This and subsequent work supported
design modifications that led to a doubling of the commercial plant capacity.
Scheme 1: DuPont’s reaction sequence for tetrahydrofuran production via selective oxidation of n-butane to maleic anhydride.
EXPERIMENTAL
Commercial plant configuration
DuPont developed CFB technology to produce maleic anhydride from n-butane with a
proprietary VPP catalyst [7, 19]. The concept was based on shuttling catalyst between two
reactor zones: VPP was reduced by n-butane in a transport bed/riser and then re-oxidized in a
fluidized bed regenerator. Transferring catalyst between the two zones was intended to maintain
the catalyst in the most favorable oxidation state to achieve higher selectivity typical of fixed bed
reactors while maintaining the superior heat transfer characteristics of fluidized bed reactors.
VPP
OOO
H2O
HOOC COOH
H2
O
n-Butane MaleicAnhydride
MaleicAcid
Tetrahydrofuran
O2CO + CO2 + H2O
Combustion
219
Lattice oxygen is more selective and active compared to molecular oxygen, thus not only are
yields better, but also, the safety hazards are minimized.
Figure 1 shows the commercial CFB facility consisting of a transport bed operating in the dense
phase fast or turbulent fluidization regime (fast bed), a riser, a stripper, and a regenerator.
Catalyst entered the base of the fast bed through a tapered standpipe and was transported upwards
to the riser. The fast bed contained two sets of cooling coils and three oxygen spargers with 926
nozzles per sparger. The riser operated at gas velocities of about 6 m/s and at catalyst suspension
densities as high as 200 kg/m3. The fast bed of the commercial reactor was 4.2 m in diameter and
11 m tall, while the riser was 1.8 m in diameter and 25 m tall. The gas-powder mixture exited the
riser and entered the stripper tangentially, which provided a sufficiently high level of separation
so that the cyclone was never overloaded. Based on iso-kinetic solids sampling in the horizontal
duct between the stripper exit and cyclone entrance, the solids mass flow rate was 0.13 t/h, which
represents a 96% stripping efficiency. The solids descended into a stripper bed equipped with
horizontal coils. They descended along a standpipe and entered the air-fed regenerator from the
top. The regenerator was equipped with horizontal cooling coils.
Pilot plant configuration
Figure 2 illustrates the sparger configuration in the fast bed of the pilot plant. During the final
demonstration period, spargers were installed at three levels, with four nozzles in the bottom, two
in the middle, and three in the top sparger. The shrouds were pointed inward at an angle of about
9° from the wall and at a 24° angle from vertical so that the oxygen would neither impinge onto
the fast bed wall nor onto the cooling coils. The shroud was 51 mm in length with an inner
diameter of 7 mm and an orifice diameter of 3.7 mm. The sparger design flow rate per shroud
was about 7 kg/h, and nominal velocities through the shrouds were 9 to 20 m/s and as high as 40
m/s. This nozzle velocity and geometry are critical to the dispersion of oxygen into the catalyst
bed. To illustrate, Figures 3a and 3b demonstrate the bubble formation of air sparged into a
beaker of water at a rate of 14 and 40 m/s, respectively. At the lower gas velocity, bubbles form
at the nozzle and detach at a constant frequency, whereas bubble formation and detachment are
more random at the higher gas velocity. Figure 3c shows bubble formation at 40 m/s around a
modified nozzle with a sintered metal filter substituted for the standard shroud. The bubbles are
220
much smaller with the sintered metal filter compared to the standard shroud, and the gas is
distributed over a larger surface area, which presumably reduces the risk of homogeneous
reactions.
Over the course of the pilot plant demonstration, the catalyst activity, as measured off-line in a
standardized test, dropped over a short period of time, although extended testing in laboratory
reactors indicated that activity was stable over several months. The drop in activity corresponded
to a decrease in the catalyst surface area. In addition, a fraction of the catalyst turned from a gray
color to black. Based on controlled thermal experiments, VPP powder became black when it was
exposed to temperatures above 700oC. Such a high temperature was never recorded anywhere in
the pilot plant even though it was well instrumented with dozens of thermocouples. Sudden
changes were observed in the local temperature in the fast bed of as much as 20°C, but more
often the temperature rise detected was on the order of 2°C. The sudden change in temperature
was accompanied by a drop in the oxygen concentration and a corresponding increase in carbon
dioxide. These local “exotherms” were at times self extinguishing but could continue for several
hours (or more).
The exotherms were attributed to localized hot spots around the sparger nozzles in which pure
oxygen was fed to the fast bed. Although the average gas-phase composition in the reactor was
outside the flammability envelope, the local composition could cross the flammability envelope
as the oxygen jet (or bubbles) was dispersed into the fast bed emulsion phase. Since the
temperature was above the autoignition temperature, the onset of combustion was determined by
the induction time of the local gas phase and solids concentration. Under mild conditions (low n-
butane feed rates), the emulsion phase effectively quenched free radicals, and the induction time
was longer than the time it takes for the oxygen to blend with the catalyst and recycle gas.
However, at high n-butane and oxygen feed rates, the induction time was lower than the mixing
time. Thus the emulsion phase was incapable of quenching free radicals, resulting in a sustained
flame. The combination of an increase in temperature and CO2 concentration and a decrease in
oxygen concentration is referred to here as a “temperature excursion.”
Laboratory Fluidized Bed Reactor Configuration
221
In the present study, a laboratory-scale bubbling fluidized bed reactor with VPP catalyst was used
to investigate the effects of temperature, pressure, mixture composition, and solids fraction on the
oxidation of n-butane/oxygen mixtures. The fluidized bed was operated in the bubbling regime
with superficial gas velocities ranging from 0.3 to 0.6 m/s with mixtures of n-butane and oxygen
in nitrogen to simulate the conditions in the fast bed, the cyclone/stripper, and the regenerator of
a CFB reactor. The main objectives were to identify the operating conditions necessary to
minimize n-butane combustion and to maximize maleic anhydride yield. In general, the fluidized
bed was operated at temperatures between 320 and 390°C and pressures from 3 to 4 bar.
Autoignition was detected by monitoring the temperature axially along the reactor while
simultaneously sampling the gas phase. The sampled gas was analyzed with a gas
chromatograph, a paramagnetic oxygen analyzer, and an FT-IR spectrometer to monitor the
conversion of the oxygen and hydrocarbon species. A detailed description of the experimental
apparatus follows.
Figure 4 shows a process flow diagram of the experimental apparatus. The process consists of a
heated reactant delivery system, a 12.8 cm inner diameter (I.D.) fluidized bed reactor with a
disengagement zone, reactor effluent and sampling lines, and various process and analytical
instrumentation. Typical operating conditions for homogeneous gas phase combustion
experiments include approximately 5 mol% n-butane, 2 mol% oxygen, 1 mol% maleic anhydride
(MAN), (balance nitrogen), and a reactor temperature and pressure of 360°C and 4 bar,
respectively.
The reactant delivery system incorporates nitrogen as the carrier gas which is metered using a
mass flow controller (MFC, Brooks Model 5853E, 500 L/min). This stream is preheated to
approximately 210°C using electric gas circulation heaters (Omega Engineering Model AHPF-
122). n-Butane is metered as a liquid from a liquefied gas cylinder equipped with a dip tube and
containing a nitrogen head pressure of about 7 bar. The n-butane cylinder mass is monitored and
recorded using an electronic balance (Mettler model ID5 Multirange Weighing Terminal
equipped with an Option 089 serial interface). A pressure regulator (PR) reduces the n-butane
pressure prior to delivery to a diaphragm liquid metering pump (Pulsafeeder Model 7120-S-AE).
A pulsation dampener and a back pressure regulator are used to maintain pressure on the pump
effluent and to dampen pressure pulsations resulting from operation of the diaphragm pump.
This liquid n-butane stream is then injected into the heated nitrogen stream and vaporized,
222
resulting in a gaseous n-butane/nitrogen stream of approximately 100°C and 4 bar. When
desired, a syringe pump (Isco Model 500D) is used to meter additional molten or liquid materials
into this reactor feed stream where they are similarly vaporized. For example, maleic anhydride
(m.p. 55°C) can be melted in situ by heating the pump cylinder using a custom heating mantle
(Glas-Col), and the molten material can then be metered into this feed mixing point. The
combined gaseous feed stream then enters a gas circulation electric heater (Caloritech Model
EX1609F467MRY, 9 kW) which raises the temperature to the desired reactor feed levels, and the
stream is then delivered to the reactor inlet. The oxygen stream is fed from a cylinder through a
two-stage gas regulator to a mass flow controller (Brooks Model 5851E, 100 L/min) before
passing through a downward pointing Type 321SS oxygen sparger into the fluidized bed section
of the reactor. The sparger is equipped with a 2.1 mm i.d. orifice and has overall dimensions of
10 mm o.d. by 5 mm i.d., with a 51 mm long shroud. This oxygen stream is heated by electrical
heating tape on the external reactor feed tubing and by the sparger which is located in the hot
reactor zone.
Figure 5 shows a cross-sectional view of the fluidized bed reactor. This approximately 46 L
pressure vessel was custom designed by DuPont and fabricated by SL Piping Systems (Newport,
Delaware). The reactor is equipped with a 0.32-cm thick, 10-micron sintered metal (316 SS) gas
distribution plate that supports the solid catalyst at the bottom of the reactor. The heated feed
mixture enters at the base of the reactor through a 1.6 cm I.D. pipe into a plenum and is then
distributed by the sintered metal disk to fluidize and contact the catalyst. The fluidized bed
portion of the reactor is constructed of 12.8 cm I.D. carbon steel pipe and has several threaded
couplings which are used for pressure and temperature instrumentation, online gas sampling, and
insertion of an oxygen sparger. This 12.8 cm section expands with a cone angle of approximately
22° to a flanged 31.9-cm I.D. section, which serves as a disengagement zone to separate the gases
from the solids in the reactor. The overall height of the reactor is 157 cm. The reactor design
specifies a maximum allowable working pressure of 6.2 bar, and the reactor was hydrostatically
proof tested to 9.6 bar prior to installation. Pressure relief is provided by a 5-cm Inconel rupture
disk (R/D, BS&B Model CSR reverse buckling type) rated at a burst pressure of 6.2 bar at 22°C.
The reactor effluent gases exit through two sintered metal pipes (Mott Model 2232-A08-12-A00-
0.5-AA, 5.1 cm O.D., 30.5 cm long, 0.5 micron porosity, 316SS) with a combined surface area of
about 930 cm2. The effluent from the two sintered metal pipes is then recombined and passes
223
through a back pressure regulator which is manipulated to maintain the desired reactor operating
pressure. The reactor is charged with a commercial VPP catalyst (Geldart A, typically about 4-5
kg) which is fluidized by the inlet gas stream.
Gas samples from various sample ports in the reactor system were analyzed by an on-line gas
chromatograph (GC). These analyses were performed using a Hewlett-Packard Model HP 5890
Series II GC equipped with both thermal conductivity (TCD) and flame ionization (FID)
detectors and a multidimensional method using four columns located in the same GC oven. The
relatively complex analysis method is a modification of that developed by Mills and Guise [20].
The TCD detector was used for analysis of oxygen, nitrogen, carbon dioxide, carbon monoxide,
water, and unreacted n-butane in packed columns. The FID detector was used for the analysis of
the hydrocarbons and oxygenates with capillary columns operated in the split mode (40:1)
utilizing a split/splitless capillary injector. The on-line gas samples were injected from a 10-port
Valco gas-sampling valve equipped with two 500-µL sample loops so that a given gas sample
could be collected and injected to both sides of the multiple column and two-detector
configuration. The FID side was equipped with a 30 m x 0.53 mm I.D. x 0.3 µm film Restek
1701 column and a 25 m x 0.53 mm I.D. Al2O3/KCl PLOT (Chrompack) column. These
columns were connected to a six-port Valco switching valve so that the helium carrier gas flows
through either the Restek 1701 column only, or through both columns in series. The TCD side
uses two packed columns: a 61 cm x 0.32 cm O.D. column packed with 80/100 mesh Hayesep R
and a 305 cm x 0.32 cm O.D. column packed with 60/80 mesh molecular sieve 5A. The analysis
utilized a temperature program of 35°C for six minutes, and then the temperature was ramped at
10°C/min to 75°C, then another temperature ramp of 20°C/min to 165°C with a final hold period
of 5.5 min. The GC was calibrated for the primary components using a certified gas mixture
consisting of 5.00% carbon dioxide, 5.00% carbon monoxide, 5.00% oxygen, 1.50 % n-butane,
balance nitrogen (MG Industries, Morrisville, PA).
Additional process equipment and instrumentation includes an on-line paramagnetic oxygen
analyzer (Siemens model OXYMAT 5F), two on-line process infrared (IR) spectrometers for
n-butane and carbon dioxide (Combustion Engineering model 501B Process IR Analyzers) and
various pressure transducer indicators and Type K thermocouples. A data acquisition system
(LabTech Notebook™, ver. 10.0, Laboratory Technologies Corporation, Wilmington, MA)
records pertinent reactor temperatures and pressures, nitrogen and oxygen flow rates, the weight
224
of the n-butane feed cylinder, and the effluent oxygen, n-butane, and carbon dioxide
concentrations from the on-line analyzers. The pressures at the top of the reactor and at the
effluent back pressure regulator are used to measure the pressure drop across the exit filters.
The reactor is primarily heated by the hot feed gases entering the base of the reactor. A heating
band is included on the upper 32-cm flange of the reactor since this was determined to be a
primary source of reactor heat losses. Typical operations include running this and the main
reactor feed heater overnight with site nitrogen flow to the reactor in order to maintain reactor
temperatures in a “hot standby” condition and hence increase experimental productivity.
The apparatus is located in a process barricade designed to safely contain a burst reactor or
auxiliary equipment. The system is equipped with an interlock system that closes the oxygen and
n-butane feed streams and shuts off the n-butane pump if an “emergency stop” button is pressed,
if a loss of barricade exhaust or nitrogen flow is sensed, or if a high pressure signal is sensed at
the primary feed heater inlet or at the top of the reactor. If a loss of feed nitrogen is sensed, an
automatic valve opens which allows a second nitrogen source to sweep the process free of any
flammable gases. In addition, the heater controllers for the primary feed heater and the top
reactor flange heater are equipped with high-temperature interlocks to prevent overheating during
unattended “hot standby” operation.
MODEL DEVELOPMENT
In parallel with laboratory experiments, an elementary step level free-radical kinetics model was
developed to describe the homogeneous, gas-phase chemistry using the Chemkin collection of
programs and well-established reaction networks and associated parameters [21, 22], comprising
440 elementary reactions and 78 species.
n-Butane oxidation is known to exhibit cool flame behavior at moderate temperatures
characterized by a negative temperature coefficient on reaction rate. This phenomenon has been
well described by Dechaux and coworkers [12], where they show that the onset and end of this
reaction regime is a function of n-butane concentration. For instance, at 10% n-butane, cool
flame behavior was shown to be important between 335–390°C. Since the commercial n-butane
oxidation reaction system operates within this general composition and temperature range, the
description of these low temperature effects was essential to the development of a useful model.
225
The model developed by Pitz and coworkers [22] has been shown to accurately capture these
phenomena and their impact on induction time.
Reaction simulations using the Chemkin model were run in several modes. The freeboard
section of the laboratory fluidized bed was simulated as a perfectly mixed gas phase system using
the experimental inlet composition to the reactor and gas residence time of the freeboard. These
simulations were run in an adiabatic mode to approximate conditions in the freeboard of the
laboratory reactor. Induction time predictions were performed using a constant pressure, batch
simulation.
RESULTS AND DISCUSSION
Commercial experience
During the commissioning of the CFB reactor, the redox operating mode—where molecular
oxygen is only fed to the regenerator, and n-butane and recycle gas are fed to the fast bed—was
tested first. The production rate was 1,500 kg/h maleic acid (MAC). Oxygen addition to the fast
bed middle sparger was the next step, and the production rate increased to 3,500 kg/h MAC. The
third step was to begin oxygen addition to the upper sparger. While ramping up the oxygen feed
rate, a high-high temperature interlock in the product gas stream downstream of the stripper shut
the plant down. Inspection of the catalyst revealed that a small fraction turned black (less than
1%).
The high-high temperature interlock was caused by gas phase combustion in the freeboard region
of the stripper/cyclone. A similar event was recorded in the pilot plant during a four hour period
the day the plant was shut down. At that moment, the exit n-butane and oxygen concentrations
were 10 and 3.7%, respectively, the pressure was 3 bar and the temperature was 399°C.
Combustion in the exit pipe of the commercial plant occurred at 391°C and with n-butane and
oxygen concentrations of only 6 and 1.6%, respectively. Maintaining oxygen concentrations at
these elevated levels was essential to achieving high production rates.
The original configuration of the commercial plant included two spargers located at heights of 1.8
m and 5.5 m above the grid, as shown in Figure 1 immediately beneath each set of cooling coils.
A third oxygen sparger was installed 0.5 m above the grid with the exact same design as the
226
original two spargers—926 nozzles distributed uniformly throughout the reactor cross-section
and directed downward at an angle of 30°. To better distribute the oxygen, further modification
was made to substitute the standard shrouds with sintered metal filters in the upper sparger.
Several attempts to feed oxygen with the standard shrouds to the upper sparger at rates greater
than 1000 kg/h were unsuccessful due to temperature excursions. Figure 3 illustrates the bubble
formation around the sintered metal: the bubbles are much smaller with the sintered metal
compared to the shrouds, and the gas is distributed over a larger surface area, which presumably
reduces the risk of exotherms. Whereas oxygen feed rates of 1000 kg/h were unsustainable to the
upper spargers with standard shrouds, feed rates of over 2600 kg/h were successfully
demonstrated with the sintered metal filters. Note that the oxygen was diluted with nitrogen in
all three spargers. As much as 1000 kg/h was continually fed to the lower and middle sparger,
and between 200 to 500 kg/h was fed to the modified upper sparger.
Pilot plant experience
Three charges of catalyst were tested during the demonstration period that lasted about 18
months. During the first charge of catalyst, a single nozzle was installed at the center of the fast
bed directly below the bottom coils about 1 m above the recycle gas sparger. Originally, the fast
bed was 3 m tall. (An additional 3 m was added just before testing the second charge of catalyst.)
The recycle gas together with fresh n-butane were introduced into the bed with a shower cap
distributor. The distributor was found eroded following the first six months of operation leaving
a single, large hole. The specific causes of this nozzle failure were not determined, but rather, the
failed distributor was replaced with a conventional six-point candelabra sparger for testing with
catalyst charges 2 and 3, and this distributor proved more durable. Despite the high oxygen flow
rates through the single point nozzle—as much as 100 kg/h—and a compromised recycle gas
sparger, only two temperature excursions were recorded with the first charge of catalyst. Two
factors may account for the low frequency of excursions: (1) the solids from the standpipe below
the fast bed may have risen into the bed as a plume, providing excellent gas-solids contacting at
the nozzle, and (2) the n-butane feed during the first charge of catalyst contained as much as 30
ppm of sulfur, whereas the sulfur content in the n-butane was as little as 1 ppm for the following
two catalyst charges. In the form of SO2, sulfur is a known flame suppressant. As shown in
227
Figure 2, additional oxygen nozzles were added to distribute the oxygen more uniformly along
the length of the fast bed. These nozzles were located near the wall and not in the center of the
bed
During testing with the second catalyst charge, over 100 exotherms were recorded. Most were of
short duration—a couple of minutes—with a temperature rise of about 1°C; the highest
temperature rise recorded was 10°C, and the longest excursion lasted 8 hours. During the
demonstration phase with the 3rd charge of catalyst, over 50 excursions were recorded, and Table
1 summarizes the conditions in the fast bed for those that were due to process conditions. (Some
excursions occurred while ramping up and others were due to instability in the oxygen feed or
solids circulation rate.) The table is organized sequentially with the time representing on stream
operation while feeding n-butane. The average temperature, T, in the lower half of the fast bed is
tabulated in the second column. At low maleic acid production rates (<20 kg/h), the average
temperature rise in the fast bed varied between 0-6oC. At high production rates (30-50 kg/h), the
temperature rise reached about 8°C in the fast bed and as high as 8°C across the riser. The
demonstration run was conducted at a riser exit temperature of 400oC. The superficial gas
velocity, Ug, assumes that all gas streams enter at the same point in the bottom of the fast bed
(recycle gas + oxygen and nitrogen from all spargers + interstitial gas carried with the solids).
High solids mass fluxes were tested early in the program, but superior n-butane utility was
achieved with mass fluxes on the order of 76 kg/m2s in the fast bed compared to 130 kg/m2s.
Excursions were recorded with as little as 7 and 5% mixed cup concentrations of n-butane and
oxygen, respectively, and the recorded temperature rise varied from 2 to 6oC. The catalyst
surface area was initially 20 m2/g, and after 1600 h of operation it dropped below 15 m2/g. The
successive drops in surface area seemed to correlate with the occurrence of exotherms.
The excursion at 5% oxygen was recorded at the lowest bed bulk density of 323 kg/m3. The
average bed density during the other excursions was close to 370 kg/m3. Based on a multi-
parameter regression analysis, feed oxygen concentration and bed suspension density were
identified as critical parameters that initiated excursions.
Together with assessing catalyst performance, identifying operating parameters to extinguish
temperature excursions became an objective of the third charge of catalyst. The combination of
high n-butane and oxygen feed rates was a prime factor that correlated with frequency and
228
severity of excursions, and maximizing these parameters was critical for the overall process
economics. Other risk factors that were identified included unstable catalyst circulation rate, high
temperature, high gas velocity, and low overall pressure drop, which translated to low solids
density. The most effective means of extinguishing an excursion was to reduce the oxygen feed
rate to the sparger. Increasing the fraction of particles below 44 µm from 16% to 23% allowed
an increase in the oxygen feed rate by 10%. Diluting the oxygen with nitrogen also was effective
at extinguishing exotherms. During the last two weeks of the demonstration run, oxygen was fed
at high rates with no temperature excursions, and at the same time, the design operating condition
targets were achieved, including the space time yield. Initially, oxygen was fed through the
bottom spargers with four shrouds. Temperature excursions were recorded with a combined total
oxygen feed rate of about 10 kg/h per nozzle with a maximum n-butane concentration of 10%.
To attempt to feed more oxygen to the reactor, the feed rate was reduced to the bottom sparger
and increased to the upper sparger. A third sparger was installed in between the two at the end of
the demonstration period. During the final stages of demonstration, total oxygen feed rates of 70
kg/h were achieved at 15% n-butane. The feed rates to the oxygen spargers were about 7.5 kg/h,
and this rate was used for the commercial design.
Fluidized Bed Reactor and Modeling Evaluations
The laboratory fluidized bed reactor apparatus and theoretical modeling were utilized to address a
number of design and operating issues related to the addition of molecular oxygen into the
commercial CFB reactor. The objectives of this experimental and modeling program were to (1)
study the local reaction behavior near an oxygen sparger placed in a fluidized bed of solid VPP
catalyst, and (2) evaluate the effect of operating conditions on the onset of combustion in the
homogeneous gas phase region of the n-butane oxidation reactor system. The fluidized bed
reactor apparatus described above was used to simulate plant operating conditions (e.g., reactor
temperature, pressure, concentrations, and solids density) and the effect of manipulating these
experimental factors on suppression of gas phase combustion. Experimental factors investigated
include these reactor conditions as well as the oxygen sparger design and operation. The primary
goals of this work were to enhance production capacity by safely maximizing the free oxygen
229
concentration in the n-butane oxidation reactor and to optimize the operating conditions and
oxygen sparger design to allow injection of the corresponding increased oxygen flow rates.
To illustrate the importance of the operating conditions on n-butane oxidation in the reactor,
Figure 6 shows the impact of relatively modest pressure perturbations in the reactor freeboard on
the oxidation chemistry occurring in the reactor. In general, the reactor pressure was held
constant at about 3.3 bar, and the inlet n-butane composition was fixed at 4%. The initial inlet
oxygen composition was about 4%, and then the oxygen flow rate was increased at about 17 min
run time and again at 22 min run time to give inlet compositions of 5 and 6 %, respectively.
Overall, the reactor temperature rose gradually during the course of this run until the second
increase in oxygen composition, at which time the temperature rose exponentially, and the run
was aborted.
The indicated pressure spikes of 0.3-0.6 bar were induced by back flushing the reactor filters with
nitrogen for about a 3-second pulse to clear the filters of accumulated catalyst fines. These
sudden spikes in the overall reactor pressure occurred because during the brief filter backflush,
half of the normal effluent flow area (i.e., the flow through one of the two sintered metal filters)
was temporarily removed from the effluent flow path, while at the same time, additional nitrogen
was added to the reactor through the filter. The combination of reduced flow area and additional
nitrogen flow into the reactor resulted in a sudden and rapid rise in pressure until the blowback
valves were diverted back to their normal setting.
As shown in Figure 6, each pressure spike in the homogeneous gas phase region of the reactor
resulted in an immediate corresponding spike in temperature from autoignition of the reaction
mixture. The indicated temperature spikes at 7.0, 11.9, 20.2, and 25.1 min run time correspond
exactly with the pressure spikes resulting from back flushing the filters. Replicate tests
confirmed this temperature spiking at similar conditions and isolated the cause to that of the
corresponding pressure spikes. These and similar experiments demonstrated that for these
particular operating conditions, autoignition can be induced by pressure spikes of as little as 0.3-
0.6 bar.
Effects of Operating Conditions within the Fluidized Bed
230
Initial experiments involved running parametric studies on the fluidized bed reactor to understand
the influence of various operating parameters on the oxidation chemistry occurring within the
fluidized catalyst section of the reactor. For these experiments, the gas-phase sampling system
was configured to monitor the effluent from the fluidized bed section of the reactor. Figure 7
shows typical results from one of these parametric studies. In particular, this plot shows the
effect of reactor temperature and inlet oxygen composition on the molecular oxygen conversion
and both CO2 and maleic anhydride (MAN) selectivity in the fluidized catalyst bed. These
selectivities are calculated on the basis of the converted oxygen. The convention of tracking
oxygen conversion, rather than n-butane conversion was selected because most conditions
explored were in the fuel rich region. As expected, for a given initial oxygen composition, more
of the inlet oxygen flow is converted with increasing temperature. However, since the selective
partial oxidation to maleic anhydride is the dominant reaction at these conditions, the conversion
of the inlet oxygen decreases with increasing excess oxygen composition. That is, the increasing
amount of excess oxygen present in the fluidized catalyst bed with increasing inlet oxygen
composition is not converted in the absence of significant homogeneous combustion. The
selectivity to the partial oxidation product maleic anhydride increases marginally with increasing
oxygen composition, but exhibits little influence of the operating temperature in this range. The
selectivity to the combustion product carbon dioxide is essentially constant at about 15% over the
range of temperatures and oxygen compositions evaluated.
During startup of the commercial reactor, approximately 25% of the bottom tier spargers and
50% of the top tier spargers developed pluggage from catalyst solids. As described above,
porous metal nozzles were considered and ultimately installed on one of the reactor spargers to
allow increased oxygen flow rates. In anticipation of this installation, a porous metal tipped
sparger similar in design to that planned for the commercial reactor was tested in the laboratory
fluidized bed reactor to determine if this modification could also produce undesired effects (e.g.,
the porous metal tending to stabilize a flame at the sparger tip). Initially, a standard sparger
similar in design to the current plant spargers (0.082” i.d. orifice with a 0.405” o.d. x 0.197” i.d. x
2” long shroud) was run at base case conditions. An identical sparger was then modified by
welding to the tip a standard 5-micron porous metal Inconel cup filter obtained from Mott
Metallurgical Corp. (Mott P.N. 1200-.500-.375-1.00-5um, Mott CPN # 1204350-17-050). This
sparger was then installed and run at similar conditions for comparison. Figure 8 shows typical
231
results comparing the two spargers. These results are plotted as molecular oxygen conversion
versus the initial reactor oxygen composition at 362°C, 4.7% initial n-butane composition, and
pressures of 3 and 4 bar. These results reflect analysis of gas samples taken from near the top of
the fluidized bed section of the reactor prior to disengagement of the solids in the reactor
freeboard. Figure 8 shows that varying the reactor pressure from 3 to 4 bar had a negligible
effect on the molecular oxygen conversion at these conditions, and combined with analytical
results for the effluent gas composition, demonstrates that complete combustion is not initiated at
the sparger under these conditions in the fluidized catalyst bed. However, Figure 8 demonstrates
a small but measureable difference in the molecular oxygen conversion obtained for the two
spargers. The porous metal sparger resulted in lower oxygen conversion at both pressures.
Similar results (not shown) indicating a small effect of sparger design were observed at the more
stringent operating conditions of 385°C, 4 bar, and 7% n-butane feed composition, which
approximated plant design conditions.
Figure 9 shows similar data using the porous metal sparger in the fluidized bed at 385°C, 4.1 bar,
and both 7% and 10% n-butane feed compositions. Again, the former concentration
approximates plant design conditions. These results and the effluent gas composition confirm
that combustion is not initiated at the sparger under these conditions.
Effects of Operating Conditions within the Reactor Freeboard
The experimental and modeling efforts were next applied to dynamic testing of the impact of
operating conditions in the reactor freeboard to simulate the low solids sections of the
commercial reactor. The goal of these efforts was to better understand conditions contributing to
homogeneous gas phase combustion in the low solids environment found in the riser stripper and
cyclone with an intent of suppressing autoignition in these zones in commercial operations.
Figure 10a shows the influence of temperature on the molecular oxygen conversion in the reactor
freeboard as a function of the initial oxygen composition. This figure plots the increase in
temperature observed in the freeboard over the course of the experiment following introduction
of oxygen at the indicated feed composition. Figure 10b shows the corresponding molecular
oxygen conversion and the combined CO and CO2 selectivity for these conditions. For the given
operating conditions of 4% n-butane, 3.3 bar, and an approximate freeboard residence time of 6.3
232
sec, the exothermic temperature rise is only 6°C at the lowest initial reactor operating
temperature shown of 320°C. The corresponding oxygen conversion and CO/CO2 selectivity
shown in Figure 10b are relatively constant with increasing molecular oxygen composition at
approximately 13 and 26%, respectively. These data suggest that selective heterogeneous
oxidation is the dominant mechanism at these conditions. Conversely, for these same operating
conditions but at the higher initial temperatures shown of 335, 357, and 380°C, a significant
exotherm is observed in the freeboard, which is combined with a dramatic increase in molecular
oxygen conversion with high selectivity to CO and CO2. These data are consistent with
homogeneous gas phase combustion.
The laboratory fluidized bed reactor was also used to study residence time effects on the onset of
autoignition in this low solids homogeneous gas phase section. The motivation for this work was
to support a planned commercial reactor modification to decrease the diameter of the transfer line
from the reactor stripper to the cyclone, which is a potential area for autoignition to occur. The
intent was to decrease the residence time of the homogeneous gas phase mixture exiting the
stripper, to let the material sweep through the system before the induction time for combustion.
Figure 11 shows the effect of residence time and initial molecular oxygen composition on the
conversion of molecular oxygen at 362°C, 4.5% initial n-butane composition, and a pressure
corresponding to the design pressure (4.1 bar) of the commercial reactor. Note that this
conversion includes the oxygen consumption in both the fluidized bed and in the freeboard, but
the freeboard residence time is estimated based on the low-solids reactor volume which excludes
the fluidized bed section. The data show that at these conditions, decreasing the "low solids" gas-
phase residence time in the freeboard delays the onset of complete combustion (as characterized
by total oxygen conversion) to higher molecular oxygen compositions. This trend is expected for
a free radical combustion chemistry mechanism with autoignition which characteristically
includes an induction period for initiation.
The freeboard reaction was modeled using a perfectly mixed reactor model in an adiabatic mode.
Initial conditions and process parameters for the model runs were selected to mimic the
experimental conditions shown in Figure 11. The initial gas composition was assumed to contain
4.5% n-butane with varying amounts of O2 ranging from 0.75–2.5% with the balance as N2.
Simulations were performed for three gas residence times, 4.5, 6.3 and 8.3 seconds, and at an
233
initial temperature of 362°C and pressure of 4.1 bar. The predicted oxygen conversion levels are
lower than those observed experimentally; however, the shape of the response with respect to O2
concentration, and the effect of increased residence time is qualitatively consistent with the
experimental results. Lower predicted conversions can be explained by two factors. First, the
model results do not reflect reactions that have occurred in the fast bed. These include both
oxygen consuming reactions, as well as heterogeneous reactions that may contribute to a source
of free radicals at the inlet to the freeboard. Based partially on these confirming experimental
data and modeling results, the stripper transfer line in the commercial reactor was replaced with a
smaller diameter pipe to decrease the residence time.
Figure 12 illustrates the pronounced effect of pressure on molecular oxygen conversion in the
freeboard of the laboratory fluidized bed reactor at 360°C, 4.5% n-butane, 6.3 s freeboard
residence time, and 3.1 and 4.1 bar—pressures corresponding to the operating pressure of the
pilot plant and the commercial plant design pressure, respectively. These data suggest that at
these conditions, increased oxygen conversion due to autoignition and combustion reactions will
occur at a significantly lower molecular oxygen concentration at the higher design pressure than
for the lower operating pressure. These results reinforce the results of Figure 6 presented
previously.
The freeboard reactor model was run at 4.5% n-butane and oxygen compositions ranging from 1
to 2.5%, with an initial temperature of 360°C for two pressures, 3.1 and 4.1 bar. Clearly, oxygen
conversion increases with increasing initial O2 levels, and higher pressure promotes the
homogeneous consumption of O2 significantly. Again, predicted conversions are lower than
what is observed experimentally for the same reasons cited above, however, qualitatively the
trends predicted by the model are consistent with experiment.
This pressure effect is further explored in Figure 13 at the 1% initial O2 level and 6.3 s residence
time, where the molecular O2 conversion is shown to increase from 42% at 2.75 bar, to nearly
88% at 4.1 bar for an initial reactor temperature of 360°C. Further increases in the reactor
pressure at these conditions resulted in complete combustion in the freeboard as indicated by
complete oxygen consumption with a simultaneous dramatic increase in the temperature and
CO/CO2 selectivity (not shown). The comparable model runs indicate very little conversion
between 2.75 and 3 bar, however, reaction does appear to take off when the initial pressure is
234
raised to 3.12 bar, and then rises in a similar fashion to the experimental results. The lack of
reaction at the lowest pressures indicates that the oxidation induction time at 1% O2 and these
pressures is longer than the time available for reaction. The fact that reaction is observed
experimentally for these conditions suggests that free radicals created heterogeneously in the
fastbed serve to seed the homogeneous reactions in the freeboard, thereby reducing induction
time.
Modeling Oxidation Induction Time
Induction time is explored directly in model runs shown in Figures 14 and 15. In Figures 14(a)
and (b), induction time in a constant pressure batch system is tested at two different initial n-
butane concentrations, 7.4 and 4.4%, at an initial temperature of 385°C and 4.1 bar, parametric in
initial O2 concentration. For the higher 7.4% initial n-butane case, the induction time for all O2
levels is on the order of about 6 s. At 4.4% n-butane, the induction time is about 14 s. The
magnitude of the pressure rise, which is indicative of the reaction rate, is amplified at higher
initial O2 concentration, but the onset of reaction occurs at approximately the same time. Figure
15 shows the important impact of pressure on oxidation induction time. Here the 13 s induction
time at 4 bar increases to 200 s at 1 bar. These results show that the onset of reaction, or
oxidation induction time, is inversely proportional to the pressure squared. These induction time
results do not map directly to the freeboard observations since the back-mixing of free radicals in
the freeboard serve to accelerate reaction initiation, however all trends are consistent.
SUMMARY AND CONCLUSIONS
DuPont’s n-butane oxidation process involves the use of a proprietary attrition-resistant VPP
catalyst and CFB reactor technology for the selective oxidation of n-butane to maleic anhydride
as an intermediate for the production of tetrahydrofuran. Reactor operating conditions fall within
the flammability envelope of the reactant and products to maximize reactor productivity. Such
operation is possible since the solids phase tends to suppress non-selective homogeneous
reactions (combustion) while at the same time promoting the desired selective catalytic reactions.
During operation of both a pilot plant and the commissioning of the commercial CFB reactor,
“temperature excursions” resulting from localized combustion reactions were experienced both in
235
the vicinity of spargers supplying oxygen to the reactors and in the homogeneous gas phase or
“free board” region in and downstream of the reactor cyclone.
A laboratory fluidized bed reactor and theoretical modeling were utilized to address design and
operating issues related to the addition of molecular oxygen into the commercial CFB reactor
with a goal of enhancing production capacity by safely maximizing the free oxygen
concentration. The simulation consisted of an elementary step level free-radical kinetics model
using the Chemkin collection of programs and well-established reaction networks and associated
parameters comprising 440 elementary reactions and 78 species. The objectives of this
experimental and modeling program were (1) to study the local reaction behavior near an oxygen
sparger placed in a fluidized bed of solid VPP catalyst, and (2) to evaluate the effect of operating
conditions on the onset of combustion in the homogeneous gas phase region of the n-butane
oxidation reactor. Plant operating conditions (e.g., reactor temperature, pressure, residence time,
and n-butane and oxygen compositions) were simulated in the fluidized bed reactor to determine
the effect of manipulating these experimental factors on suppression of gas phase combustion.
In general, the laboratory fluidized bed reactor was operated at temperatures between 320 and
390°C, pressures from 3 to 4 bar, n-butane compositions of 4-10%, oxygen compositions of 1-
6%, and freeboard residence times of 4-8 s. Parametric studies of the oxidation products are
presented for both the fluidized bed and freeboard sections of the reactor. Both experimental and
modeling results demonstrated a pronounced pressure effect on the onset of combustion in this
range of operating conditions. In one example, multiple temperature excursions corresponding to
autoignition were induced by small pressure spikes of as little as 0.3-0.6 bar. A significant
impact of operating temperature on the onset of combustion was also noted between 320 and
335°C in the reactor freeboard, with combustion being the dominant oxidation mechanism at
335°C and higher temperatures. Both experimental and modeling results demonstrated that
decreasing the residence time in the low solids regions of the reactor delays the onset of complete
combustion to longer time or higher relative molecular oxygen compositions. Kinetic model
predictions are presented that show a significant influence of reactant composition and reactor
pressure on oxidation induction time. Finally, the effect of using porous metal oxygen spargers
as alternatives to standard nozzles was also demonstrated. Little impact of this sparger design on
inducing combustion in the fluidized bed was observed over a range of operating conditions.
236
Subsequent papers will include reports of additional work on mitigating combustion by
quenching effluent gases from the high solids sections of the reactor, the effect of suspended
catalyst fines on suppressing autoignition, the effect of the product MAN composition, and the
impact of adding free radical scavengers to suppress combustion.
REFERENCES
[1] P. Botella, J.M. López Nieto, B. Solsona, J. Mol. Catal. A: Chem. 184 (2002) 335-347.
[2] M.J. Lorences, G.S. Patience, R. Cenni, F. Diez, J. Coca, Catal. Today 112 (2006) 45-48.
[3] T.V. Malleswara Rao, G. Deo, Ind. Eng. Chem. Res. 46 (2007) 70-79.
[4] R.P. Hesketh, J.F. Davidson, Combust. Flame 85 (1991) 449-467.
[5] D.R. van der Vaart, Fuel 67 (1988) 1003-1007.
[6] N. Ballarini, F. Cavani, C. Cortelli, S. Ligi, F. Pierelli, F. Trifiro, C. Fumagalli, G. Mazzoni, T. Monti, Topics Catal. 38 (2006) 147-156.
[7] R.M. Contractor, Chem. Eng. Sci. 54 (1999) 5627-5632.
[8] M. Cathonnet, J.C. Boettner, H. James, Eighteenth Symposium (International) on Combustion, The Combustion Institute, 1981, pp. 903-913.
[9] A. Burcat, K. Scheller, A. Lifshitz, Combust. Flame 16 (1971) 29-33.
[10] S. Kojima, T. Suzuoki, Combust. Flame 92 (1993) 254-265.
[11] R. Minetti, M. Ribaucour, M. Carlier, C. Fittschen, L.R. Sochet, Combust. Flame 96 (1994) 201-211.
[12] J.C. Déchaux, J.L. Flament, M. Lucquin, Combust. Flame 17 (1971) 205-214.
[13] G.A. Luckett, R.T. Pollard, Combust. Flame 21 (1973) 265-274.
[14] M.B. Neumann, P.M. Tutakin, Compt. Rend. Acad. Sci. URSS 4 (1936) 127-130.
[15] D.C. Horning, D.F. Davidson, R.K. Hanson, J. Propul. Power 18 (2002) 363-371.
[16] M.A. Oehlschlaeger, D.F. Davidson, J.T. Herbon, R.K. Hanson, Int. J. Chem. Kinet. 36 (2004) 67-78.
[17] R.J. Olsen, W.R. Williams, X. Song, L.D. Schmidt, R. Aris, Chem. Eng. Sci. 47 (1992) 2505-2510.
[18] M. Ziauddin, G. Veser, L.D. Schmidt, Catal. Lett. 46 (1997) 159-167.
[19] G.S. Patience, R.E. Bockrath, Appl. Catal., A, in press, this issue.
[20] P.L. Mills, W.E. Guise, Jr., J. Chromatogr. Sci. 34 (1996) 431-459.
[21] W.J. Pitz, C.K. Westbrook, Combust. Flame 63 (1986) 113-133.
[22] W.J. Pitz, R.D. Wilk, C.K. Westbrook, N.P. Cernansky, Western States Section/The Combustion Institute, Salt Lake City, UT, Paper No. WSSCI 88-51 (1988).
237
List of Tables
Table 1. Pilot Plant Temperature Excursions
238
Table 1. Pilot Plant Temperature Excursions
Time
(h)
T
(oC)
Gas Velocity
(m/s)
Solids Mass Flux (kg/m2s)
Pressure Drop
(mbar)
Suspension Density (kg/m3)
[n-C4H10]
(mol%)
[O2]
(mol%)
Bottom Sparger (kg/h)
Top Sparger (kg/h)
Middle Sparger (kg/h)
BET S.A.
(m2/g)
∆T (oC)
83 372 1.1 98 618 379 7.6 11.1 45 0 0 20.2 2
85 371 1.3 115 640 366 7.3 9.5 41 0 0 20.1 2
97 375 1.3 117 616 347 8.7 8.5 41 0 0 19.9 2
98 378 1.3 117 612 345 10.0 7.8 41 0 0 19.9 2
99 377 1.3 117 612 345 10.0 8.0 41 0 0 19.9 2
100 377 1.3 116 614 345 8.0 8.6 41 0 0 19.9 2
102 376 1.3 117 616 345 9.2 8.4 41 0 0 19.9 4
105 374 1.3 113 629 363 10.0 8.2 40 0 0 19.8 4
107 370 1.3 116 644 372 10.9 7.9 39 0 0 19.8 4
109 372 1.3 126 652 364 10.3 8.3 40 0 0 19.8 2
137 375 1.2 116 646 370 10.6 9.0 35 8 0 19.3 3
141 374 1.2 114 642 369 10.6 9.8 38 8 0 19.3 2
144 369 1.2 114 654 376 11.1 9.7 36 7 0 19.2 2
146 371 1.2 115 640 376 10.7 9.6 36 7 0 19.2 2
207 371 1.3 121 638 365 11.1 7.3 36 0 0 18.3 2
213 370 1.2 123 675 387 10.9 10.3 36 9 0 18.2 2
239
214 371 1.2 124 673 387 10.9 10.1 36 9 0 18.2 2
216 374 1.2 123 669 382 10.6 9.9 36 9 0 18.2 2
217 372 1.2 124 672 387 10.8 9.7 35 9 0 18.1 2
222 370 1.2 127 682 392 10.1 10.0 35 8 0 18.1 2
574 374 1.5 116 559 324 15.1 4.4 0 0 0 17.4 4
579 373 1.5 121 566 323 14.1 5.7 31 0 0 17.4 2
607 381 1.3 129 694 416 13.8 9.0 31 0 0 17.4 2
1000 383 1.3 112 639 361 11.7 12.7 25 0 16 16.6 3
1004 382 1.3 111 637 360 11.6 13.9 27 0 30 16.6 4
1248 381 0.9 76 680 449 17.4 17.4 30 14 30 16.1 4
1525 383 0.9 61 621 338 15.5 15.3 30 14 30 15.1 3
1526 382 0.9 61 624 347 15.7 15.0 27 14 29 15.1 6
1552 384 0.9 61 602 340 15.1 15.6 25 10 24 14.8 4
240
List of Figures
Figure 1. Circulating Fluidized Bed (CFB) configuration for the commercial facility.
Figure 2. Pilot plant reactor fast bed configuration.
Figure 3. Sparger tests with air sparged through two nozzle geometries: Standard nozzle in water at (a) 14 m/s, (b) 40 m/s; (c) Sintered metal nozzle in ethanol at 40 m/s.
Figure 4. Experimental fluidized bed reactor apparatus.
Figure 5. Cross-sectional view of the fluidized bed reactor.
Figure 6. Pressure effect showing reactor freeboard temperature excursions induced by temporary reactor pressure spikes.
Figure 7. Effect of reactor temperature and initial oxygen composition on molecular oxygen conversion and CO2 and maleic anhydride (MAN) selectivity in the fluidized bed: O2 conversion at • 362°C, □ 375°C, � 387°C; CO2 selectivity at ○ 362°C, ■ 375°C, � 387°C; MAN selectivity at � 362°C, � 375°C, � 387°C.
Figure 8. Influence of nozzle type and reactor pressure on molecular oxygen conversion in the fluidized bed.
Figure 9. n-Butane concentration effect in fluidized bed with porous metal sparger.
Figure 10. Influence of reactor temperature and initial oxygen composition on molecular oxygen conversion and CO/CO2 selectivity in the reactor freeboard: (a) reactor temperature rise at • 380°C, □ 357°C, � 335°C, ∆ 320°C; (b) O2 conversion at ○ 380°C, □ 357°C, � 335°C, ∆ 320°C; CO/CO2 selectivity at • 380°C, ■ 357°C, � 335°C, ▲ 320°C.
Figure 11. Effect of freeboard residence time and inlet oxygen composition on molecular oxygen conversion (lines in experimental data are to guide the eye).
Figure 12. Effect of reactor pressure and inlet oxygen composition on molecular oxygen conversion in the reactor freeboard.
Figure 13. Effect of reactor pressure on molecular oxygen conversion in the reactor freeboard.
Figure 14. Effect of initial oxygen composition on induction time – constant pressure batch simulation: (a) 7.4% initial n-butane concentration (b) 4.4% initial n-butane concentration.
Figure 15. Model predictions of the effect of pressure on oxidation induction time: (a) Temperature rise vs time, parametric in pressure (b) Induction time vs P.
241
Figures
242
Figure 1. Circulating Fluidized Bed (CFB) configuration for the commercial facility.
Reactor
Cyclone
Regenerator
Fast
Bed
Riser
Stripper
243
Figure 2. Pilot plant reactor fast bed configuration.
Figure 3. Sparger tests with air sparged through two nozzle geometries: Standard nozzle in
water at (a) 14 m/s, (b) 40 m/s; (c) Sintered metal nozzle in ethanol at 40 m/s.
(a)
(b)
(c)
Figure 3. Sparger tests with air sparged through two nozzle geometries: Standard nozzle in
m/s, (b) 40 m/s; (c) Sintered metal nozzle in ethanol at 40 m/s.
244
Figure 3. Sparger tests with air sparged through two nozzle geometries: Standard nozzle in
m/s, (b) 40 m/s; (c) Sintered metal nozzle in ethanol at 40 m/s.
245
Figure 4. Experimental fluidized bed reactor apparatus.
Figure 5. Cross
Figure 5. Cross-sectional view of the fluidized bed reactor.
246
247
Figure 6. Pressure effect showing reactor freeboard temperature excursions induced by
temporary reactor pressure spikes.
2.5
3.0
3.5
4.0
4.5
5.0
0 5 10 15 20 25 30 35
Run Time (min)
Re
ac
tor
Pre
ss
ure
(b
ar)
0
100
200
300
400
500
Re
acto
r T
em
pe
ratu
re (
oC
)
[n-C4H10] = 4%
[O2] = 4-6%
7.0 min
11.9 min20.2 min
25.1 min
248
Figure 7. Effect of reactor temperature and initial oxygen composition on molecular oxygen
conversion and CO2 and maleic anhydride (MAN) selectivity in the fluidized catalyst bed: O2
conversion at • 362°C, □ 375°C, � 387°C; CO2 selectivity at ○ 362°C, ■ 375°C, � 387°C;
MAN selectivity at � 362°C, � 375°C, � 387°C.
0
10
20
30
40
50
60
70
80
90
100
0 1 2 3 4
Initial O2 Composition (mol%)
Co
nve
rsio
n o
r S
ele
cti
vit
y (
%)
[n-C4H10] = 5%
P = 4.1 bar
249
Figure 8. Influence of nozzle type and reactor pressure on molecular oxygen conversion
in the fluidized catalyst bed.
40
50
60
70
80
90
100
0 1 2 3 4
Initial O2 Composition (mol%)
O2 C
on
vers
ion
(%
)
3 bar, Standard 3 bar, Porous Metal 4 bar, Standard 4 bar, Porous Metal
[n-C4H10] = 4.7%
T = 362°C
P Nozzle
250
Figure 9. n-Butane concentration effect in fluidized bed with porous metal sparger.
60
65
70
75
80
85
90
95
100
0 1 2 3 4
Initial O2 Composition (mol%)
O2 C
on
ve
rsio
n (
%)
7%
10%
T = 385°C
P = 4.1 bar
[n-C4H10]
251
(a)
(b)
Figure 10. Influence of reactor temperature and initial oxygen composition on molecular oxygen conversion and CO/CO2 selectivity in the reactor freeboard: (a) reactor temperature rise at
• 380°C, □ 357°C, � 335°C, ∆ 320°C; (b) O2 conversion at ○ 380°C, □ 357°C, � 335°C, ∆ 320°C; CO/CO2 selectivity at • 380°C, ■ 357°C, � 335°C, ▲ 320°C.
0
20
40
60
80
100
0 1 2 3 4 5
Initial O2 Composition (mol%)
Re
ac
tor
Te
mp
era
ture
Ris
e (
oC
)
[n-C4H10] = 4%
P = 3.3. bar
τ = 6 s
0
20
40
60
80
100
0 1 2 3 4 5
Initial O2 Composition (mol%)
Co
nvers
ion
or
Sele
cti
vit
y (
%)
[n-C4H10] = 4%
P = 3.3. bar
τ = 6 s
Figure 11. Effect of freeboard residence time and inlet oxygen composition on molecular oxygen
conversion (lines in experimental data are to guide the eye).
Figure 11. Effect of freeboard residence time and inlet oxygen composition on molecular oxygen
conversion (lines in experimental data are to guide the eye).
252
Figure 11. Effect of freeboard residence time and inlet oxygen composition on molecular oxygen
253
Figure 12. Effect of reactor pressure and inlet oxygen composition on molecular oxygen
conversion in the reactor freeboard.
0
20
40
60
80
100
0.5 1.0 1.5 2.0 2.5 3.0
Initial O2 Composition (mol%)
O2 C
on
vers
ion
(%
)
4.1 bar, Model3.1 bar, Model4.1 bar, Expt.3.1 bar, Expt.
T = 360°C
[n-C4H10] = 4.5%
τ = 6.3 s
254
Figure 13. Effect of reactor pressure on molecular oxygen conversion in the reactor freeboard.
0
10
20
30
40
50
60
70
80
90
100
2.0 2.5 3.0 3.5 4.0 4.5 5.0
Reactor Pressure (bar)
O2 C
on
vers
ion
(%
)
Expt
Model
T = 360 °C
[n-C4H10] = 4.4%
[O2] = 1.0%
255
(a)
(b)
Figure 14. Effect of initial oxygen composition on induction time – constant pressure batch simulation: (a) 7.4% initial n-butane concentration (b) 4.4% initial n-butane concentration.
350
400
450
500
550
600
0 10 20 30 40 50 60 70 80
time [s]
T [
°C]
0.5
1
1.5
2
2.5
3
[n-C4H10]=7.4%
P=4.1 bar
[%O2]
350
400
450
500
550
600
650
0 10 20 30 40 50 60 70 80
time [s]
T [
°C]
0.5
1
1.5
2
2.5
3
[n-C4H10]=4.4%
P=4.1 bar
[%O2]
256
(a)
(b)
Figure 15. Model predictions of the effect of pressure on oxidation induction time: (a) Temperature rise vs time, parametric in pressure (b) Induction time vs P.
350
400
450
500
550
600
650
0 50 100 150 200 250 300
time [s]
T [
°C]
1
2
3
4
P [bar]
0
50
100
150
200
0 1 2 3 4 5
Pressure [bar]
Ind
uc
tio
n T
ime
[s
]
2200Ind Time [s] =
P
257
ANNEXE 9 – BREVET POUR LA MÉTHODE DE MESURE SIMULTANÉE
DE LA FRACTION DE SOLIDES ET DE LA COMPOSITION CHIMIQUE
DE LA PHASE GAZEUSE
258
259